Temporally Coherent Embeddings for Self-Supervised Video Representation Learning

Joshua Knights, Ben Harwood, Daniel Ward, Anthony Vanderkop, Olivia Mackenzie-Ross, Peyman Moghadam

Responsive image

Auto-TLDR; Temporally Coherent Embeddings for Self-supervised Video Representation Learning

Slides Poster

This paper presents TCE: Temporally Coherent Embeddings for self-supervised video representation learning. The proposed method exploits inherent structure of unlabeled video data to explicitly enforce temporal coherency in the embedding space, rather than indirectly learning it through ranking or predictive proxy tasks. In the same way that high-level visual information in the world changes smoothly, we believe that nearby frames in learned representations will benefit from demonstrating similar properties. Using this assumption, we train our TCE model to encode videos such that adjacent frames exist close to each other and videos are separated from one another. Using TCE we learn robust representations from large quantities of unlabeled video data. We thoroughly analyse and evaluate our self-supervised learned TCE models on a downstream task of video action recognition using multiple challenging benchmarks (Kinetics400, UCF101, HMDB51). With a simple but effective 2D-CNN backbone and only RGB stream inputs, TCE pre-trained representations outperform all previous self-supervised 2D-CNN and 3D-CNN trained on UCF101. The code and pre-trained models for this paper can be downloaded at: https://github.com/csiro-robotics/TCE

Similar papers

Audio-Visual Predictive Coding for Self-Supervised Visual Representation Learning

Mani Kumar Tellamekala, Michel Valstar, Michael Pound, Timo Giesbrecht

Responsive image

Auto-TLDR; AV-PPC: A Multi-task Learning Framework for Learning Semantic Visual Features from Unlabeled Video Data

Slides Poster Similar

Self-supervised learning has emerged as a candidate approach to learn semantic visual features from unlabeled video data. In self-supervised learning, intrinsic correspondences between data points are used to define a proxy task that forces the model to learn semantic representations. Most existing proxy tasks applied to video data exploit only either intra-modal (e.g. temporal) or cross-modal (e.g. audio-visual) correspondences separately. In theory, jointly learning both these correspondences may result in richer visual features; but, as we show in this work, doing so is non-trivial in practice. To address this problem, we introduce `Audio-Visual Permutative Predictive Coding' (AV-PPC), a multi-task learning framework designed to fully leverage the temporal and cross-modal correspondences as natural supervision signals. In AV-PPC, the model is trained to simultaneously learn multiple intra- and cross-modal predictive coding sub-tasks. By using visual speech recognition (lip-reading) as the downstream evaluation task, we show that our proposed proxy task can learn higher quality visual features than existing proxy tasks. We also show that AV-PPC visual features are highly data-efficient. Without further finetuning, AV-PPC visual encoder achieves 80.30% spoken word classification rate on the LRW dataset, performing on par with directly or fully supervised visual encoders learned from large amounts of labeled data.

Self-Supervised Joint Encoding of Motion and Appearance for First Person Action Recognition

Mirco Planamente, Andrea Bottino, Barbara Caputo

Responsive image

Auto-TLDR; A Single Stream Architecture for Egocentric Action Recognition from the First-Person Point of View

Slides Poster Similar

Wearable cameras are becoming more and more popular in several applications, increasing the interest of the research community in developing approaches for recognizing actions from the first-person point of view. An open challenge in egocentric action recognition is that videos lack detailed information about the main actor's pose and thus tend to record only parts of the movement when focusing on manipulation tasks. Thus, the amount of information about the action itself is limited, making crucial the understanding of the manipulated objects and their context. Many previous works addressed this issue with two-stream architectures, where one stream is dedicated to modeling the appearance of objects involved in the action, and another to extracting motion features from optical flow. In this paper, we argue that learning features jointly from these two information channels is beneficial to capture the spatio-temporal correlations between the two better. To this end, we propose a single stream architecture able to do so, thanks to the addition of a self-supervised block that uses a pretext motion prediction task to intertwine motion and appearance knowledge. Experiments on several publicly available databases show the power of our approach.

RMS-Net: Regression and Masking for Soccer Event Spotting

Matteo Tomei, Lorenzo Baraldi, Simone Calderara, Simone Bronzin, Rita Cucchiara

Responsive image

Auto-TLDR; An Action Spotting Network for Soccer Videos

Slides Poster Similar

The recently proposed action spotting task consists in finding the exact timestamp in which an event occurs. This task fits particularly well for soccer videos, where events correspond to salient actions strictly defined by soccer rules (a goal occurs when the ball crosses the goal line). In this paper, we devise a lightweight and modular network for action spotting, which can simultaneously predict the event label and its temporal offset using the same underlying features. We enrich our model with two training strategies: the first one for data balancing and uniform sampling, the second for masking ambiguous frames and keeping the most discriminative visual cues. When tested on the SoccerNet dataset and using standard features, our full proposal exceeds the current state of the art by 3 Average-mAP points. Additionally, it reaches a gain of more than 10 Average-mAP points on the test set when fine-tuned in combination with a strong 2D backbone.

Not 3D Re-ID: Simple Single Stream 2D Convolution for Robust Video Re-Identification

Toby Breckon, Aishah Alsehaim

Responsive image

Auto-TLDR; ResNet50-IBN for Video-based Person Re-Identification using Single Stream 2D Convolution Network

Slides Poster Similar

Video-based person re-identification has received increasing attention recently, as it plays an important role within the surveillance video analysis. Video-based Re-ID is an expansion of earlier image-based re-identification methods by learning features from a video via multiple image frames for each person. Most contemporary video Re-ID methods utilise complex CNN-based network architectures using 3D convolution or multi-branch networks to extract spatial-temporal features from the video. By contrast, in this paper, we will illustrate superior performance from a simple single stream 2D convolution network leveraging the ResNet50-IBN architecture to extract frame-level features followed by temporal attention for clip level features. These clip level features can be generalised to extract video level features by averaging clip level features without any additional cost. Our model, uses best video Re-ID practice and transfer learning between datasets, outperforms existing state-of-the-art approaches on MARS, PRID2011 and iLIDSVID datasets with 89:62%, 97:75%, 97:33% rank-1 accuracy respectively and with 84:61% mAP for MARS, without reliance on complex and memory intensive 3D convolutions or multistream networks architectures as found in other contemporary work. Conversely, this work shows that global features extracted by the 2D convolution network are a sufficient representation for robust state of the art video Re-ID.

The Color Out of Space: Learning Self-Supervised Representations for Earth Observation Imagery

Stefano Vincenzi, Angelo Porrello, Pietro Buzzega, Marco Cipriano, Pietro Fronte, Roberto Cuccu, Carla Ippoliti, Annamaria Conte, Simone Calderara

Responsive image

Auto-TLDR; Satellite Image Representation Learning for Remote Sensing

Slides Poster Similar

The recent growth in the number of satellite images fosters the development of effective deep-learning techniques for Remote Sensing (RS). However, their full potential is untapped due to the lack of large annotated datasets. Such a problem is usually countered by fine-tuning a feature extractor that is previously trained on the ImageNet dataset. Unfortunately, the domain of natural images differs from the RS one, which hinders the final performance. In this work, we propose to learn meaningful representations from satellite imagery, leveraging its high-dimensionality spectral bands to reconstruct the visible colors. We conduct experiments on land cover classification (BigEarthNet) and West Nile Virus detection, showing that colorization is a solid pretext task for training a feature extractor. Furthermore, we qualitatively observe that guesses based on natural images and colorization rely on different parts of the input. This paves the way to an ensemble model that eventually outperforms both the above-mentioned techniques.

Learnable Higher-Order Representation for Action Recognition

Jie Shao, Xiangyang Xue

Responsive image

Auto-TLDR; Learningable Higher-Order Operations for Spatiotemporal Dynamics in Video Recognition

Similar

Capturing spatiotemporal dynamics is an essential topic in video recognition. In this paper, we present learnable higher-order operations as a generic family of building blocks for capturing spatiotemporal dynamics from RGB input video space. Similar to higher-order functions, the weights of higher-order operations are themselves derived from the data with learnable parameters. Classical architectures such as residual learning and network-in-network are first-order operations where weights are directly learned from the data. Higher-order operations make it easier to capture context-sensitive patterns, such as motion. Self-attention models are also higher-order operations, but the attention weights are mostly computed from an affine operation or dot product. The learnable higher-order operations can be more generic and flexible. Experimentally, we show that on the task of video recognition, our higher-order models can achieve results on par with or better than the existing state-of-the-art methods on Something-Something (V1 and V2), Kinetics and Charades datasets.

Self-Supervised Learning of Dynamic Representations for Static Images

Siyang Song, Enrique Sanchez, Linlin Shen, Michel Valstar

Responsive image

Auto-TLDR; Facial Action Unit Intensity Estimation and Affect Estimation from Still Images with Multiple Temporal Scale

Slides Poster Similar

Facial actions are spatio-temporal signals by nature, and therefore their modeling is crucially dependent on the availability of temporal information. In this paper, we focus on inferring such temporal dynamics of facial actions when no explicit temporal information is available, i.e. from still images. We present a novel approach to capture multiple scales of such temporal dynamics, with an application to facial Action Unit (AU) intensity estimation and dimensional affect estimation. In particular, 1) we propose a framework that infers a dynamic representation (DR) from a still image, which captures the bi-directional flow of time within a short time-window centered at the input image; 2) we show that we can train our method without the need of explicitly generating target representations, allowing the network to represent dynamics more broadly; and 3) we propose to apply a multiple temporal scale approach that infers DRs for different window lengths (MDR) from a still image. We empirically validate the value of our approach on the task of frame ranking, and show how our proposed MDR attains state of the art results on BP4D for AU intensity estimation and on SEMAINE for dimensional affect estimation, using only still images at test time.

Multi-Level Deep Learning Vehicle Re-Identification Using Ranked-Based Loss Functions

Eleni Kamenou, Jesus Martinez-Del-Rincon, Paul Miller, Patricia Devlin - Hill

Responsive image

Auto-TLDR; Multi-Level Re-identification Network for Vehicle Re-Identification

Slides Poster Similar

Identifying vehicles across a network of cameras with non-overlapping fields of view remains a challenging research problem due to scene occlusions, significant inter-class similarity and intra-class variability. In this paper, we propose an end-to-end multi-level re-identification network that is capable of successfully projecting same identity vehicles closer to one another in the embedding space, compared to vehicles of different identities. Robust feature representations are obtained by combining features at multiple levels of the network. As for the learning process, we employ a recent state-of-the-art structured metric learning loss function previously applied to other retrieval problems and adjust it to the vehicle re-identification task. Furthermore, we explore the cases of image-to-image, image-to-video and video-to-video similarity metric. Finally, we evaluate our system and achieve great performance on two large-scale publicly available datasets, CityFlow-ReID and VeRi-776. Compared to most existing state-of-art approaches, our approach is simpler and more straightforward, utilizing only identity-level annotations, while avoiding post-processing the ranking results (re-ranking) at the testing phase.

Attention-Based Deep Metric Learning for Near-Duplicate Video Retrieval

Kuan-Hsun Wang, Chia Chun Cheng, Yi-Ling Chen, Yale Song, Shang-Hong Lai

Responsive image

Auto-TLDR; Attention-based Deep Metric Learning for Near-duplicate Video Retrieval

Slides Similar

Near-duplicate video retrieval (NDVR) is an important and challenging problem due to the increasing amount of videos uploaded to the Internet. In this paper, we propose an attention-based deep metric learning method for NDVR. Our method is based on well-established principles: We leverage two-stream networks to combine RGB and optical flow features, and incorporate an attention module to effectively deal with distractor frames commonly observed in near duplicate videos. We further aggregate the features corresponding to multiple video segments to enhance the discriminative power. The whole system is trained using a deep metric learning objective with a Siamese architecture. Our experiments show that the attention module helps eliminate redundant and noisy frames, while focusing on visually relevant frames for solving NVDR. We evaluate our approach on recent large-scale NDVR datasets, CC_WEB_VIDEO, VCDB, FIVR and SVD. To demonstrate the generalization ability of our approach, we report results in both within- and cross-dataset settings, and show that the proposed method significantly outperforms state-of-the-art approaches.

SSDL: Self-Supervised Domain Learning for Improved Face Recognition

Samadhi Poornima Kumarasinghe Wickrama Arachchilage, Ebroul Izquierdo

Responsive image

Auto-TLDR; Self-supervised Domain Learning for Face Recognition in unconstrained environments

Slides Poster Similar

Face recognition in unconstrained environments is challenging due to variations in illumination, quality of sensing, motion blur and etc. An individual’s face appearance can vary drastically under different conditions creating a gap between train (source) and varying test (target) data. The domain gap could cause decreased performance levels in direct knowledge transfer from source to target. Despite fine-tuning with domain specific data could be an effective solution, collecting and annotating data for all domains is extremely expensive. To this end, we propose a self-supervised domain learning (SSDL) scheme that trains on triplets mined from unlabelled data. A key factor in effective discriminative learning, is selecting informative triplets. Building on most confident predictions, we follow an “easy-to-hard” scheme of alternate triplet mining and self-learning. Comprehensive experiments on four different benchmarks show that SSDL generalizes well on different domains.

What and How? Jointly Forecasting Human Action and Pose

Yanjun Zhu, Yanxia Zhang, Qiong Liu, Andreas Girgensohn

Responsive image

Auto-TLDR; Forecasting Human Actions and Motion Trajectories with Joint Action Classification and Pose Regression

Slides Poster Similar

Forecasting human actions and motion trajectories addresses the problem of predicting what a person is going to do next and how they will perform it. This is crucial in a wide range of applications such as assisted living and future co-robotic settings. We propose to simultaneously learn actions and action-related human motion dynamics, while existing works perform them independently. In this paper, we present a method to jointly forecast categories of human action and the pose of skeletal joints in the hope that the two tasks can help each other. As a result, our system can predict not only the future actions but also the motion trajectories that will result. To achieve this, we define a task of joint action classification and pose regression. We employ a sequence to sequence encoder-decoder model combined with multi-task learning to forecast future actions and poses progressively before the action happens. Experimental results on two public datasets, IkeaDB and OAD, demonstrate the effectiveness of the proposed method.

Single View Learning in Action Recognition

Gaurvi Goyal, Nicoletta Noceti, Francesca Odone

Responsive image

Auto-TLDR; Cross-View Action Recognition Using Domain Adaptation for Knowledge Transfer

Slides Poster Similar

Viewpoint is an essential aspect of how an action is visually perceived, with the motion appearing substantially different for some viewpoint pairs. Data driven action recognition algorithms compensate for this by including a variety of viewpoints in their training data, adding to the cost of data acquisition as well as training. We propose a novel methodology that leverages deeply pretrained features to learn actions from a single viewpoint using domain adaptation for knowledge transfer. We demonstrate the effectiveness of this pipeline on 3 different datasets: IXMAS, MoCA and NTU RGBD+, and compare with both classical and deep learning methods. Our method requires low training data and demonstrates unparalleled cross-view action recognition accuracies for single view learning.

Feature-Supervised Action Modality Transfer

Fida Mohammad Thoker, Cees Snoek

Responsive image

Auto-TLDR; Cross-Modal Action Recognition and Detection in Non-RGB Video Modalities by Learning from Large-Scale Labeled RGB Data

Slides Poster Similar

This paper strives for action recognition and detection in video modalities like RGB, depth maps or 3D-skeleton sequences when only limited modality-specific labeled examples are available. For the RGB, and derived optical-flow, modality many large-scale labeled datasets have been made available. They have become the de facto pre-training choice when recognizing or detecting new actions from RGB datasets that have limited amounts of labeled examples available. Unfortunately, large-scale labeled action datasets for other modalities are unavailable for pre-training. In this paper, our goal is to recognize actions from limited examples in non-RGB video modalities, by learning from large-scale labeled RGB data. To this end, we propose a two-step training process: (i) we extract action representation knowledge from an RGB-trained teacher network and adapt it to a non-RGB student network. (ii) we then fine-tune the transfer model with available labeled examples of the target modality. For the knowledge transfer we introduce feature-supervision strategies, which rely on unlabeled pairs of two modalities (the RGB and the target modality) to transfer feature level representations from the teacher to the the student network. Ablations and generalizations with two RGB source datasets and two non-RGB target datasets demonstrate that an optical-flow teacher provides better action transfer features than RGB for both depth maps and 3D-skeletons, even when evaluated on a different target domain, or for a different task. Compared to alternative cross-modal action transfer methods we show a good improvement in performance especially when labeled non-RGB examples to learn from are scarce.

Nonlinear Ranking Loss on Riemannian Potato Embedding

Byung Hyung Kim, Yoonje Suh, Honggu Lee, Sungho Jo

Responsive image

Auto-TLDR; Riemannian Potato for Rank-based Metric Learning

Slides Poster Similar

We propose a rank-based metric learning method by leveraging a concept of the Riemannian Potato for better separating non-linear data. By exploring the geometric properties of Riemannian manifolds, the proposed loss function optimizes the measure of dispersion using the distribution of Riemannian distances between a reference sample and neighbors and builds a ranked list according to the similarities. We show the proposed function can learn a hypersphere for each class, preserving the similarity structure inside it on Riemannian manifold. As a result, compared with Euclidean distance-based metric, our method can further jointly reduce the intra-class distances and enlarge the inter-class distances for learned features, consistently outperforming state-of-the-art methods on three widely used non-linear datasets.

MFI: Multi-Range Feature Interchange for Video Action Recognition

Sikai Bai, Qi Wang, Xuelong Li

Responsive image

Auto-TLDR; Multi-range Feature Interchange Network for Action Recognition in Videos

Slides Poster Similar

Short-range motion features and long-range dependencies are two complementary and vital cues for action recognition in videos, but it remains unclear how to efficiently and effectively extract these two features. In this paper, we propose a novel network to capture these two features in a unified 2D framework. Specifically, we first construct a Short-range Temporal Interchange (STI) block, which contains a Channels-wise Temporal Interchange (CTI) module for encoding short-range motion features. Then a Graph-based Regional Interchange (GRI) module is built to present long-range dependencies using graph convolution. Finally, we replace original bottleneck blocks in the ResNet with STI blocks and insert several GRI modules between STI blocks, to form a Multi-range Feature Interchange (MFI) Network. Practically, extensive experiments are conducted on three action recognition datasets (i.e., Something-Something V1, HMDB51, and UCF101), which demonstrate that the proposed MFI network achieves impressive results with very limited computing cost.

Developing Motion Code Embedding for Action Recognition in Videos

Maxat Alibayev, David Andrea Paulius, Yu Sun

Responsive image

Auto-TLDR; Motion Embedding via Motion Codes for Action Recognition

Slides Poster Similar

We propose a motion embedding strategy via the motion codes that is a vectorized representation of motions based on their salient mechanical attributes. We show that our motion codes can provide robust motion representation. We train a deep neural network model that learns to embed demonstration videos into motion codes. We integrate the extracted features from the motion embedding model into the current state-of-the-art action recognition model. The obtained model achieved higher accuracy than the baseline on a verb classification task from egocentric videos in EPIC-KITCHENS dataset.

A Grid-Based Representation for Human Action Recognition

Soufiane Lamghari, Guillaume-Alexandre Bilodeau, Nicolas Saunier

Responsive image

Auto-TLDR; GRAR: Grid-based Representation for Action Recognition in Videos

Slides Poster Similar

Human action recognition (HAR) in videos is a fundamental research topic in computer vision. It consists mainly in understanding actions performed by humans based on a sequence of visual observations. In recent years, HAR have witnessed significant progress, especially with the emergence of deep learning models. However, most of existing approaches for action recognition rely on information that is not always relevant for the task, and are limited in the way they fuse temporal information. In this paper, we propose a novel method for human action recognition that encodes efficiently the most discriminative appearance information of an action with explicit attention on representative pose features, into a new compact grid representation. Our GRAR (Grid-based Representation for Action Recognition) method is tested on several benchmark datasets that demonstrate that our model can accurately recognize human actions, despite intra-class appearance variations and occlusion challenges.

Learning Embeddings for Image Clustering: An Empirical Study of Triplet Loss Approaches

Kalun Ho, Janis Keuper, Franz-Josef Pfreundt, Margret Keuper

Responsive image

Auto-TLDR; Clustering Objectives for K-means and Correlation Clustering Using Triplet Loss

Slides Poster Similar

In this work, we evaluate two different image clustering objectives, k-means clustering and correlation clustering, in the context of Triplet Loss induced feature space embeddings. Specifically, we train a convolutional neural network to learn discriminative features by optimizing two popular versions of the Triplet Loss in order to study their clustering properties under the assumption of noisy labels. Additionally, we propose a new, simple Triplet Loss formulation, which shows desirable properties with respect to formal clustering objectives and outperforms the existing methods. We evaluate all three Triplet loss formulations for K-means and correlation clustering on the CIFAR-10 image classification dataset.

Multi-Modal Deep Clustering: Unsupervised Partitioning of Images

Guy Shiran, Daphna Weinshall

Responsive image

Auto-TLDR; Multi-Modal Deep Clustering for Unlabeled Images

Slides Poster Similar

The clustering of unlabeled raw images is a daunting task, which has recently been approached with some success by deep learning methods. Here we propose an unsupervised clustering framework, which learns a deep neural network in an end-to-end fashion, providing direct cluster assignments of images without additional processing. Multi-Modal Deep Clustering (MMDC), trains a deep network to align its image embeddings with target points sampled from a Gaussian Mixture Model distribution. The cluster assignments are then determined by mixture component association of image embeddings. Simultaneously, the same deep network is trained to solve an additional self-supervised task. This pushes the network to learn more meaningful image representations and stabilizes the training. Experimental results show that MMDC achieves or exceeds state-of-the-art performance on four challenging benchmarks. On natural image datasets we improve on previous results with significant margins of up to 11% absolute accuracy points, yielding an accuracy of 70% on CIFAR-10 and 61% on STL-10.

SL-DML: Signal Level Deep Metric Learning for Multimodal One-Shot Action Recognition

Raphael Memmesheimer, Nick Theisen, Dietrich Paulus

Responsive image

Auto-TLDR; One-Shot Action Recognition using Metric Learning

Slides Similar

Recognizing an activity with a single reference sample using metric learning approaches is a promising research field. The majority of few-shot methods focus on object recognition or face-identification. We propose a metric learning approach to reduce the action recognition problem to a nearest neighbor search in embedding space. We encode signals into images and extract features using a deep residual CNN. Using triplet loss, we learn a feature embedding. The resulting encoder transforms features into an embedding space in which closer distances encode similar actions while higher distances encode different actions. Our approach is based on a signal level formulation and remains flexible across a variety of modalities. It further outperforms the baseline on the large scale NTU RGB+D 120 dataset for the One-Shot action recognition protocol by \ntuoneshotimpro%. With just 60% of the training data, our approach still outperforms the baseline approach by \ntuoneshotimproreduced%. With 40% of the training data, our approach performs comparably well as the second follow up. Further, we show that our approach generalizes well in experiments on the UTD-MHAD dataset for inertial, skeleton and fused data and the Simitate dataset for motion capturing data. Furthermore, our inter-joint and inter-sensor experiments suggest good capabilities on previously unseen setups.

TinyVIRAT: Low-Resolution Video Action Recognition

Ugur Demir, Yogesh Rawat, Mubarak Shah

Responsive image

Auto-TLDR; TinyVIRAT: A Progressive Generative Approach for Action Recognition in Videos

Slides Poster Similar

The existing research in action recognition is mostly focused on high-quality videos where the action is distinctly visible. In real-world surveillance environments, the actions in videos are captured at a wide range of resolutions. Most activities occur at a distance with a small resolution and recognizing such activities is a challenging problem. In this work, we focus on recognizing tiny actions in videos. We introduce a benchmark dataset, TinyVIRAT, which contains natural low-resolution activities. The actions in TinyVIRAT videos have multiple labels and they are extracted from surveillance videos which makes them realistic and more challenging. We propose a novel method for recognizing tiny actions in videos which utilizes a progressive generative approach to improve the quality of low-resolution actions. The proposed method also consists of a weakly trained attention mechanism which helps in focusing on the activity regions in the video. We perform extensive experiments to benchmark the proposed TinyVIRAT dataset and observe that the proposed method significantly improves the action recognition performance over baselines. We also evaluate the proposed approach on synthetically resized action recognition datasets and achieve state-of-the-art results when compared with existing methods. The dataset and code will be publicly available.

Generalized Local Attention Pooling for Deep Metric Learning

Carlos Roig Mari, David Varas, Issey Masuda, Juan Carlos Riveiro, Elisenda Bou-Balust

Responsive image

Auto-TLDR; Generalized Local Attention Pooling for Deep Metric Learning

Slides Poster Similar

Deep metric learning has been key to recent advances in face verification and image retrieval amongst others. These systems consist on a feature extraction block (extracts feature maps from images) followed by a spatial dimensionality reduction block (generates compact image representations from the feature maps) and an embedding generation module (projects the image representation to the embedding space). While research on deep metric learning has focused on improving the losses for the embedding generation module, the dimensionality reduction block has been overlooked. In this work, we propose a novel method to generate compact image representations which uses local spatial information through an attention mechanism, named Generalized Local Attention Pooling (GLAP). This method, instead of being placed at the end layer of the backbone, is connected at an intermediate level, resulting in lower memory requirements. We assess the performance of the aforementioned method by comparing it with multiple dimensionality reduction techniques, demonstrating the importance of using attention weights to generate robust compact image representations. Moreover, we compare the performance of multiple state-of-the-art losses using the standard deep metric learning system against the same experiment with our GLAP. Experiments showcase that the proposed Generalized Local Attention Pooling mechanism outperforms other pooling methods when compared with current state-of-the-art losses for deep metric learning.

Learning Group Activities from Skeletons without Individual Action Labels

Fabio Zappardino, Tiberio Uricchio, Lorenzo Seidenari, Alberto Del Bimbo

Responsive image

Auto-TLDR; Lean Pose Only for Group Activity Recognition

Similar

To understand human behavior we must not just recognize individual actions but model possibly complex group activity and interactions. Hierarchical models obtain the best results in group activity recognition but require fine grained individual action annotations at the actor level. In this paper we show that using only skeletal data we can train a state-of-the art end-to-end system using only group activity labels at the sequence level. Our experiments show that models trained without individual action supervision perform poorly. On the other hand we show that pseudo-labels can be computed from any pre-trained feature extractor with comparable final performance. Finally our carefully designed lean pose only architecture shows highly competitive results versus more complex multimodal approaches even in the self-supervised variant.

Graph-Based Interpolation of Feature Vectors for Accurate Few-Shot Classification

Yuqing Hu, Vincent Gripon, Stéphane Pateux

Responsive image

Auto-TLDR; Transductive Learning for Few-Shot Classification using Graph Neural Networks

Slides Poster Similar

In few-shot classification, the aim is to learn models able to discriminate classes using only a small number of labeled examples. In this context, works have proposed to introduce Graph Neural Networks (GNNs) aiming at exploiting the information contained in other samples treated concurrently, what is commonly referred to as the transductive setting in the literature. These GNNs are trained all together with a backbone feature extractor. In this paper, we propose a new method that relies on graphs only to interpolate feature vectors instead, resulting in a transductive learning setting with no additional parameters to train. Our proposed method thus exploits two levels of information: a) transfer features obtained on generic datasets, b) transductive information obtained from other samples to be classified. Using standard few-shot vision classification datasets, we demonstrate its ability to bring significant gains compared to other works.

Image Representation Learning by Transformation Regression

Xifeng Guo, Jiyuan Liu, Sihang Zhou, En Zhu, Shihao Dong

Responsive image

Auto-TLDR; Self-supervised Image Representation Learning using Continuous Parameter Prediction

Slides Poster Similar

Self-supervised learning is a thriving research direction since it can relieve the burden of human labeling for machine learning by seeking for supervision from data instead of human annotation. Although demonstrating promising performance in various applications, we observe that the existing methods usually model the auxiliary learning tasks as classification tasks with finite discrete labels, leading to insufficient supervisory signals, which in turn restricts the representation quality. In this paper, to solve the above problem and make full use of the supervision from data, we design a regression model to predict the continuous parameters of a group of transformations, i.e., image rotation, translation, and scaling. Surprisingly, this naive modification stimulates tremendous potential from data and the resulting supervisory signal has largely improved the performance of image representation learning. Extensive experiments on four image datasets, including CIFAR10, CIFAR100, STL10, and SVHN, indicate that our proposed algorithm outperforms the state-of-the-art unsupervised learning methods by a large margin in terms of classification accuracy. Crucially, we find that with our proposed training mechanism as an initialization, the performance of the existing state-of-the-art classification deep architectures can be preferably improved.

Self-Supervised Learning for Astronomical Image Classification

Ana Martinazzo, Mateus Espadoto, Nina S. T. Hirata

Responsive image

Auto-TLDR; Unlabeled Astronomical Images for Deep Neural Network Pre-training

Slides Poster Similar

In Astronomy, a huge amount of image data is generated daily by photometric surveys, which scan the sky to collect data from stars, galaxies and other celestial objects. In this paper, we propose a technique to leverage unlabeled astronomical images to pre-train deep convolutional neural networks, in order to learn a domain-specific feature extractor which improves the results of machine learning techniques in setups with small amounts of labeled data available. We show that our technique produces results which are in many cases better than using ImageNet pre-training.

Towards Practical Compressed Video Action Recognition: A Temporal Enhanced Multi-Stream Network

Bing Li, Longteng Kong, Dongming Zhang, Xiuguo Bao, Di Huang, Yunhong Wang

Responsive image

Auto-TLDR; TEMSN: Temporal Enhanced Multi-Stream Network for Compressed Video Action Recognition

Slides Poster Similar

Current compressed video action recognition methods are mainly based on completely received compressed videos. However, in real transmission, the compressed video packets are usually disorderly received and lost due to network jitters or congestion. It is of great significance to recognize actions in early phases with limited packets, e.g. forecasting the potential risks from videos quickly. In this paper, we proposed a Temporal Enhanced Multi-Stream Network (TEMSN) for practical compressed video action recognition. First, we use three compressed modalities as complementary cues and build a multi-stream network to capture the rich information from compressed video packets. Second, we design a temporal enhanced module based on Encoder-Decoder structure applied on each stream to infer the missing packets, and generate more complete action dynamics. Thanks to the rich modalities and temporal enhancement, our approach is able to better modeling the action with limited compressed packets. Experiments on HMDB-51 and UCF-101 dataset validate its effectiveness and efficiency.

Unsupervised Co-Segmentation for Athlete Movements and Live Commentaries Using Crossmodal Temporal Proximity

Yasunori Ohishi, Yuki Tanaka, Kunio Kashino

Responsive image

Auto-TLDR; A guided attention scheme for audio-visual co-segmentation

Slides Poster Similar

Audio-visual co-segmentation is a task to extract segments and regions corresponding to specific events on unlabelled audio and video signals. It is particularly important to accomplish it in an unsupervised way, since it is generally very difficult to manually label all the objects and events appearing in audio-visual signals for supervised learning. Here, we propose to take advantage of temporal proximity of corresponding audio and video entities included in the signals. For this purpose, we newly employ a guided attention scheme to this task to efficiently detect and utilize temporal cooccurrences of audio and video information. The experiments using a real TV broadcasting of Sumo wrestling, a sport event, with live commentaries show that our model can automatically extract specific athlete movements and its spoken descriptions in an unsupervised manner.

Gabriella: An Online System for Real-Time Activity Detection in Untrimmed Security Videos

Mamshad Nayeem Rizve, Ugur Demir, Praveen Praveen Tirupattur, Aayush Jung Rana, Kevin Duarte, Ishan Rajendrakumar Dave, Yogesh Rawat, Mubarak Shah

Responsive image

Auto-TLDR; Gabriella: A Real-Time Online System for Activity Detection in Surveillance Videos

Slides Similar

Activity detection in surveillance videos is a difficult problem due to multiple factors such as large field of view, presence of multiple activities, varying scales and viewpoints, and its untrimmed nature. The existing research in activity detection is mainly focused on datasets, such as UCF-101, JHMDB, THUMOS, and AVA, which partially address these issues. The requirement of processing the surveillance videos in real-time makes this even more challenging. In this work we propose Gabriella, a real-time online system to perform activity detection on untrimmed surveillance videos. The proposed method consists of three stages: tubelet extraction, activity classification, and online tubelet merging. For tubelet extraction, we propose a localization network which takes a video clip as input and spatio-temporally detects potential foreground regions at multiple scales to generate action tubelets. We propose a novel Patch-Dice loss to handle large variations in actor size. Our online processing of videos at a clip level drastically reduces the computation time in detecting activities. The detected tubelets are assigned activity class scores by the classification network and merged together using our proposed Tubelet-Merge Action-Split (TMAS) algorithm to form the final action detections. The TMAS algorithm efficiently connects the tubelets in an online fashion to generate action detections which are robust against varying length activities. We perform our experiments on the VIRAT and MEVA (Multiview Extended Video with Activities) datasets and demonstrate the effectiveness of the proposed approach in terms of speed ($\sim$100 fps) and performance with state-of-the-art results. The code and models will be made publicly available.

Building Computationally Efficient and Well-Generalizing Person Re-Identification Models with Metric Learning

Vladislav Sovrasov, Dmitry Sidnev

Responsive image

Auto-TLDR; Cross-Domain Generalization in Person Re-identification using Omni-Scale Network

Slides Similar

This work considers the problem of domain shift in person re-identification.Being trained on one dataset, a re-identification model usually performs much worse on unseen data. Partially this gap is caused by the relatively small scale of person re-identification datasets (compared to face recognition ones, for instance), but it is also related to training objectives. We propose to use the metric learning objective, namely AM-Softmax loss, and some additional training practices to build well-generalizing, yet, computationally efficient models. We use recently proposed Omni-Scale Network (OSNet) architecture combined with several training tricks and architecture adjustments to obtain state-of-the art results in cross-domain generalization problem on a large-scale MSMT17 dataset in three setups: MSMT17-all->DukeMTMC, MSMT17-train->Market1501 and MSMT17-all->Market1501.

Dual-MTGAN: Stochastic and Deterministic Motion Transfer for Image-To-Video Synthesis

Fu-En Yang, Jing-Cheng Chang, Yuan-Hao Lee, Yu-Chiang Frank Wang

Responsive image

Auto-TLDR; Dual Motion Transfer GAN for Convolutional Neural Networks

Slides Poster Similar

Generating videos with content and motion variations is a challenging task in computer vision. While the recent development of GAN allows video generation from latent representations, it is not easy to produce videos with particular content of motion patterns of interest. In this paper, we propose Dual Motion Transfer GAN (Dual-MTGAN), which takes image and video data as inputs while learning disentangled content and motion representations. Our Dual-MTGAN is able to perform deterministic motion transfer and stochastic motion generation. Based on a given image, the former preserves the input content and transfers motion patterns observed from another video sequence, and the latter directly produces videos with plausible yet diverse motion patterns based on the input image. The proposed model is trained in an end-to-end manner, without the need to utilize pre-defined motion features like pose or facial landmarks. Our quantitative and qualitative results would confirm the effectiveness and robustness of our model in addressing such conditioned image-to-video tasks.

Supervised Domain Adaptation Using Graph Embedding

Lukas Hedegaard, Omar Ali Sheikh-Omar, Alexandros Iosifidis

Responsive image

Auto-TLDR; Domain Adaptation from the Perspective of Multi-view Graph Embedding and Dimensionality Reduction

Slides Poster Similar

Getting deep convolutional neural networks to perform well requires a large amount of training data. When the available labelled data is small, it is often beneficial to use transfer learning to leverage a related larger dataset (source) in order to improve the performance on the small dataset (target). Among the transfer learning approaches, domain adaptation methods assume that distributions between the two domains are shifted and attempt to realign them. In this paper, we consider the domain adaptation problem from the perspective of multi-view graph embedding and dimensionality reduction. Instead of solving the generalised eigenvalue problem to perform the embedding, we formulate the graph-preserving criterion as loss in the neural network and learn a domain-invariant feature transformation in an end-to-end fashion. We show that the proposed approach leads to a powerful Domain Adaptation framework which generalises the prior methods CCSA and d-SNE, and enables simple and effective loss designs; an LDA-inspired instantiation of the framework leads to performance on par with the state-of-the-art on the most widely used Domain Adaptation benchmarks, Office31 and MNIST to USPS datasets.

Local Clustering with Mean Teacher for Semi-Supervised Learning

Zexi Chen, Benjamin Dutton, Bharathkumar Ramachandra, Tianfu Wu, Ranga Raju Vatsavai

Responsive image

Auto-TLDR; Local Clustering for Semi-supervised Learning

Slides Similar

The Mean Teacher (MT) model of Tarvainen and Valpola has shown favorable performance on several semi-supervised benchmark datasets. MT maintains a teacher model's weights as the exponential moving average of a student model's weights and minimizes the divergence between their probability predictions under diverse perturbations of the inputs. However, MT is known to suffer from confirmation bias, that is, reinforcing incorrect teacher model predictions. In this work, we propose a simple yet effective method called Local Clustering (LC) to mitigate the effect of confirmation bias. In MT, each data point is considered independent of other points during training; however, data points are likely to be close to each other in feature space if they share similar features. Motivated by this, we cluster data points locally by minimizing the pairwise distance between neighboring data points in feature space. Combined with a standard classification cross-entropy objective on labeled data points, the misclassified unlabeled data points are pulled towards high-density regions of their correct class with the help of their neighbors, thus improving model performance. We demonstrate on semi-supervised benchmark datasets SVHN and CIFAR-10 that adding our LC loss to MT yields significant improvements compared to MT and performance comparable to the state of the art in semi-supervised learning.

Mutual Information Based Method for Unsupervised Disentanglement of Video Representation

Aditya Sreekar P, Ujjwal Tiwari, Anoop Namboodiri

Responsive image

Auto-TLDR; MIPAE: Mutual Information Predictive Auto-Encoder for Video Prediction

Slides Poster Similar

Video Prediction is an interesting and challenging task of predicting future frames from a given set context frames that belong to a video sequence. Video prediction models have found prospective applications in Maneuver Planning, Health care, Autonomous Navigation and Simulation. One of the major challenges in future frame generation is due to the high dimensional nature of visual data. In this work, we propose Mutual Information Predictive Auto-Encoder (MIPAE) framework, that reduces the task of predicting high dimensional video frames by factorising video representations into content and low dimensional pose latent variables that are easy to predict. A standard LSTM network is used to predict these low dimensional pose representations. Content and the predicted pose representations are decoded to generate future frames. Our approach leverages the temporal structure of the latent generative factors of a video and a novel mutual information loss to learn disentangled video representations. We also propose a metric based on mutual information gap (MIG) to quantitatively access the effectiveness of disentanglement on DSprites and MPI3D-real datasets. MIG scores corroborate with the visual superiority of frames predicted by MIPAE. We also compare our method quantitatively on evaluation metrics LPIPS, SSIM and PSNR.

Reducing the Variance of Variational Estimates of Mutual Information by Limiting the Critic's Hypothesis Space to RKHS

Aditya Sreekar P, Ujjwal Tiwari, Anoop Namboodiri

Responsive image

Auto-TLDR; Mutual Information Estimation from Variational Lower Bounds Using a Critic's Hypothesis Space

Slides Similar

Mutual information (MI) is an information-theoretic measure of dependency between two random variables. Several methods to estimate MI, from samples of two random variables with unknown underlying probability distributions have been proposed in the literature. Recent methods realize parametric probability distributions or critic as a neural network to approximate unknown density ratios. The approximated density ratios are used to estimate different variational lower bounds of MI. While these methods provide reliable estimation when the true MI is low, they produce high variance estimates in cases of high MI. We argue that the high variance characteristic is due to the uncontrolled complexity of the critic's hypothesis space. In support of this argument, we use the data-driven Rademacher complexity of the hypothesis space associated with the critic's architecture to analyse generalization error bound of variational lower bound estimates of MI. In the proposed work, we show that it is possible to negate the high variance characteristics of these estimators by constraining the critic's hypothesis space to Reproducing Hilbert Kernel Space (RKHS), which corresponds to a kernel learned using Automated Spectral Kernel Learning (ASKL). By analysing the aforementioned generalization error bounds, we augment the overall optimisation objective with effective regularisation term. We empirically demonstrate the efficacy of this regularization in enforcing proper bias variance tradeoff on four variational lower bounds, namely NWJ, MINE, JS and SMILE.

Temporal Binary Representation for Event-Based Action Recognition

Simone Undri Innocenti, Federico Becattini, Federico Pernici, Alberto Del Bimbo

Responsive image

Auto-TLDR; Temporal Binary Representation for Gesture Recognition

Slides Poster Similar

In this paper we present an event aggregation strategy to convert the output of an event camera into frames processable by traditional Computer Vision algorithms. The proposed method first generates sequences of intermediate binary representations, which are then losslessly transformed into a compact format by simply applying a binary-to-decimal conversion. This strategy allows us to encode temporal information directly into pixel values, which are then interpreted by deep learning models. We apply our strategy, called Temporal Binary Representation, to the task of Gesture Recognition, obtaining state of the art results on the popular DVS128 Gesture Dataset. To underline the effectiveness of the proposed method compared to existing ones, we also collect an extension of the dataset under more challenging conditions on which to perform experiments.

A Self-Supervised GAN for Unsupervised Few-Shot Object Recognition

Khoi Nguyen, Sinisa Todorovic

Responsive image

Auto-TLDR; Self-supervised Few-Shot Object Recognition with a Triplet GAN

Slides Poster Similar

This paper addresses unsupervised few-shot object recognition, where all training images are unlabeled, and test images are divided into queries and a few labeled support images per object class of interest. The training and test images do not share object classes. We extend the vanilla GAN with two loss functions, both aimed at self-supervised learning. The first is a reconstruction loss that enforces the discriminator to reconstruct the probabilistically sampled latent code which has been used for generating the "fake" image. The second is a triplet loss that enforces the discriminator to output image encodings that are closer for more similar images. Evaluation, comparisons, and detailed ablation studies are done in the context of few-shot classification. Our approach significantly outperforms the state of the art on the Mini-Imagenet and Tiered-Imagenet datasets.

Generative Latent Implicit Conditional Optimization When Learning from Small Sample

Idan Azuri, Daphna Weinshall

Responsive image

Auto-TLDR; GLICO: Generative Latent Implicit Conditional Optimization for Small Sample Learning

Slides Poster Similar

We revisit the long-standing problem of learning from small sample. The generation of new samples from a small training set of labeled points has attracted increased attention in recent years. In this paper, we propose a novel such method called GLICO (Generative Latent Implicit Conditional Optimization). GLICO learns a mapping from the training examples to a latent space and a generator that generates images from vectors in the latent space. Unlike most recent work, which rely on access to large amounts of unlabeled data, GLICO does not require access to any additional data other than the small set of labeled points. In fact, GLICO learns to synthesize completely new samples for every class using as little as 5 or 10 examples per class, with as few as 10 such classes and no data from unknown classes. GLICO is then used to augment the small training set while training a classifier on the small sample. To this end, our proposed method samples the learned latent space using spherical interpolation (slerp) and generates new examples using the trained generator. Empirical results show that the new sampled set is diverse enough, leading to improvement in image classification in comparison with the state of the art when trained on small samples obtained from CIFAR-10, CIFAR-100, and CUB-200.

ActionSpotter: Deep Reinforcement Learning Framework for Temporal Action Spotting in Videos

Guillaume Vaudaux-Ruth, Adrien Chan-Hon-Tong, Catherine Achard

Responsive image

Auto-TLDR; ActionSpotter: A Reinforcement Learning Algorithm for Action Spotting in Video

Slides Poster Similar

Action spotting has recently been proposed as an alternative to action detection and key frame extraction. However, the current state-of-the-art method of action spotting requires an expensive ground truth composed of the search sequences employed by human annotators spotting actions - a critical limitation. In this article, we propose to use a reinforcement learning algorithm to perform efficient action spotting using only the temporal segments from the action detection annotations, thus opening an interesting solution for video understanding. Experiments performed on THUMOS14 and ActivityNet datasets show that the proposed method, named ActionSpotter, leads to good results and outperforms state-of-the-art detection outputs redrawn for this application. In particular, the spotting mean Average Precision on THUMOS14 is significantly improved from 59.7% to 65.6% while skipping 23% of video.

Video Representation Fusion Network For Multi-Label Movie Genre Classification

Tianyu Bi, Dmitri Jarnikov, Johan Lukkien

Responsive image

Auto-TLDR; A Video Representation Fusion Network for Movie Genre Classification

Slides Poster Similar

In this paper, we introduce a Video Representation Fusion Network (VRFN) for movie genre classification. Different from the previous works, which use frame-level features for movie genre classification, our approach uses video classification architecture to create video-level features from a group of frames and fuse these features temporally to learn long-term spatiotemporal information for the movie genre classification task. We use a pre-trained I3D model to generate intermediate video representations and connect it with a C3D-LSTM model for feature fusion and movie genre classification. LMTD-9 dataset which contains 4007 trailers multi-labeled with 9 movie genres is used for training and evaluation of the model. The experimental results demonstrate that learning long-term temporal dependencies by fusing video representations improves the performance in movie genre classification. Our best model outperforms the state-of-the-art methods by 3.4% improvement in AUPRC (macro).

Modeling Long-Term Interactions to Enhance Action Recognition

Alejandro Cartas, Petia Radeva, Mariella Dimiccoli

Responsive image

Auto-TLDR; A Hierarchical Long Short-Term Memory Network for Action Recognition in Egocentric Videos

Slides Poster Similar

In this paper, we propose a new approach to understand actions in egocentric videos that exploit the semantics of object interactions at both frame and temporal levels. At the frame level, we use a region-based approach that takes as input a primary region roughly corresponding to the user hands and a set of secondary regions potentially corresponding to the interacting objects and calculates the action score through a CNN formulation. This information is then fed to a Hierarchical Long Short-Term Memory Network (HLSTM) that captures temporal dependencies between actions within and across shots. Ablation studies thoroughly validate the proposed approach, showing in particular that both levels of the HLSTM architecture contribute to performance improvement. Furthermore, quantitative comparisons show that the proposed approach outperforms the state-of-the-art in terms of action recognition on standard benchmarks, without relying on motion information.

3D Attention Mechanism for Fine-Grained Classification of Table Tennis Strokes Using a Twin Spatio-Temporal Convolutional Neural Networks

Pierre-Etienne Martin, Jenny Benois-Pineau, Renaud Péteri, Julien Morlier

Responsive image

Auto-TLDR; Attentional Blocks for Action Recognition in Table Tennis Strokes

Slides Poster Similar

The paper addresses the problem of recognition of actions in video with low inter-class variability such as Table Tennis strokes. Two stream, "twin" convolutional neural networks are used with 3D convolutions both on RGB data and optical flow. Actions are recognized by classification of temporal windows. We introduce 3D attention modules and examine their impact on classification efficiency. In the context of the study of sportsmen performances, a corpus of the particular actions of table tennis strokes is considered. The use of attention blocks in the network speeds up the training step and improves the classification scores up to 5% with our twin model. We visualize the impact on the obtained features and notice correlation between attention and player movements and position. Score comparison of state-of-the-art action classification method and proposed approach with attentional blocks is performed on the corpus. Proposed model with attention blocks outperforms previous model without them and our baseline.

Video Semantic Segmentation Using Deep Multi-View Representation Learning

Akrem Sellami, Salvatore Tabbone

Responsive image

Auto-TLDR; Deep Multi-view Representation Learning for Video Object Segmentation

Slides Poster Similar

In this paper, we propose a deep learning model based on deep multi-view representation learning, to address the video object segmentation task. The proposed model emphasizes the importance of the inherent correlation between video frames and incorporates a multi-view representation learning based on deep canonically correlated autoencoders. The multi-view representation learning in our model provides an efficient mechanism for capturing inherent correlations by jointly extracting useful features and learning better representation into a joint feature space, i.e., shared representation. To increase the training data and the learning capacity, we train the proposed model with pairs of video frames, i.e., $F_{a}$ and $F_{b}$. During the segmentation phase, the deep canonically correlated autoencoders model encodes useful features by processing multiple reference frames together, which is used to detect the frequently reappearing. Our model enhances the state-of-the-art deep learning-based methods that mainly focus on learning discriminative foreground representations over appearance and motion. Experimental results over two large benchmarks demonstrate the ability of the proposed method to outperform competitive approaches and to reach good performances, in terms of semantic segmentation.

Future Urban Scenes Generation through Vehicles Synthesis

Alessandro Simoni, Luca Bergamini, Andrea Palazzi, Simone Calderara, Rita Cucchiara

Responsive image

Auto-TLDR; Predicting the Future of an Urban Scene with a Novel View Synthesis Paradigm

Slides Poster Similar

In this work we propose a deep learning pipeline to predict the visual future appearance of an urban scene. Despite recent advances, generating the entire scene in an end-to-end fashion is still far from being achieved. Instead, here we follow a two stages approach, where interpretable information is included in the loop and each actor is modelled independently. We leverage a per-object novel view synthesis paradigm; i.e. generating a synthetic representation of an object undergoing a geometrical roto-translation in the 3D space. Our model can be easily conditioned with constraints (e.g. input trajectories) provided by state-of-the-art tracking methods or by the user itself. This allows us to generate a set of diverse realistic futures starting from the same input in a multi-modal fashion. We visually and quantitatively show the superiority of this approach over traditional end-to-end scene-generation methods on CityFlow, a challenging real world dataset.

Towards Robust Learning with Different Label Noise Distributions

Diego Ortego, Eric Arazo, Paul Albert, Noel E O'Connor, Kevin Mcguinness

Responsive image

Auto-TLDR; Distribution Robust Pseudo-Labeling with Semi-supervised Learning

Slides Similar

Noisy labels are an unavoidable consequence of labeling processes and detecting them is an important step towards preventing performance degradations in Convolutional Neural Networks. Discarding noisy labels avoids a harmful memorization, while the associated image content can still be exploited in a semi-supervised learning (SSL) setup. Clean samples are usually identified using the small loss trick, i.e. they exhibit a low loss. However, we show that different noise distributions make the application of this trick less straightforward and propose to continuously relabel all images to reveal a discriminative loss against multiple distributions. SSL is then applied twice, once to improve the clean-noisy detection and again for training the final model. We design an experimental setup based on ImageNet32/64 for better understanding the consequences of representation learning with differing label noise distributions and find that non-uniform out-of-distribution noise better resembles real-world noise and that in most cases intermediate features are not affected by label noise corruption. Experiments in CIFAR-10/100, ImageNet32/64 and WebVision (real-world noise) demonstrate that the proposed label noise Distribution Robust Pseudo-Labeling (DRPL) approach gives substantial improvements over recent state-of-the-art. Code will be made available.

Shape Consistent 2D Keypoint Estimation under Domain Shift

Levi Vasconcelos, Massimiliano Mancini, Davide Boscaini, Barbara Caputo, Elisa Ricci

Responsive image

Auto-TLDR; Deep Adaptation for Keypoint Prediction under Domain Shift

Slides Poster Similar

Recent unsupervised domain adaptation methods based on deep architectures have shown remarkable performance not only in traditional classification tasks but also in more complex problems involving structured predictions (e.g. semantic segmentation, depth estimation). Following this trend, in this paper we present a novel deep adaptation framework for estimating keypoints under \textit{domain shift}, i.e. when the training (\textit{source}) and the test (\textit{target}) images significantly differ in terms of visual appearance. Our method seamlessly combines three different components: feature alignment, adversarial training and self-supervision. Specifically, our deep architecture leverages from domain-specific distribution alignment layers to perform target adaptation at the feature level. Furthermore, a novel loss is proposed which combines an adversarial term for ensuring aligned predictions in the output space and a geometric consistency term which guarantees coherent predictions between a target sample and its perturbed version. Our extensive experimental evaluation conducted on three publicly available benchmarks shows that our approach outperforms state-of-the-art domain adaptation methods in the 2D keypoint prediction task.

AttendAffectNet: Self-Attention Based Networks for Predicting Affective Responses from Movies

Thi Phuong Thao Ha, Bt Balamurali, Herremans Dorien, Roig Gemma

Responsive image

Auto-TLDR; AttendAffectNet: A Self-Attention Based Network for Emotion Prediction from Movies

Slides Poster Similar

In this work, we propose different variants of the self-attention based network for emotion prediction from movies, which we call AttendAffectNet. We take both audio and video into account and incorporate the relation among multiple modalities by applying self-attention mechanism in a novel manner into the extracted features for emotion prediction. We compare it to the typically temporal integration of the self-attention based model, which in our case, allows to capture the relation of temporal representations of the movie while considering the sequential dependencies of emotion responses. We demonstrate the effectiveness of our proposed architectures on the extended COGNIMUSE dataset [1], [2] and the MediaEval 2016 Emotional Impact of Movies Task [3], which consist of movies with emotion annotations. Our results show that applying the self-attention mechanism on the different audio-visual features, rather than in the time domain, is more effective for emotion prediction. Our approach is also proven to outperform state-of-the-art models for emotion prediction.

Precise Temporal Action Localization with Quantified Temporal Structure of Actions

Chongkai Lu, Ruimin Li, Hong Fu, Bin Fu, Yihao Wang, Wai Lun Lo, Zheru Chi

Responsive image

Auto-TLDR; Action progression networks for temporal action detection

Slides Poster Similar

Existing temporal action detection algorithms cannot distinguish complete and incomplete actions while this property is essential in many applications. To tackle this challenge, we proposed the action progression networks (APN), a novel model that predicts action progression of video frames with continuous numbers. Using the progression sequence of test video, on the top of the APN, a complete action searching algorithm (CAS) was designed to detect complete actions only. With the usage of frame-level fine-grained temporal structure modeling and detecting actions according to their whole temporal context, our framework can locate actions precisely and is good at avoiding incomplete action detection. We evaluated our framework on a new dataset (DFMAD-70) collected by ourselves which contains both complete and incomplete actions. Our framework got good temporal localization results with 95.77% average precision when the IoU threshold is 0.5. On the benchmark THUMOS14, an incomplete-ignostic dataset, our framework still obtain competitive performance. The code is available online at https://github.com/MakeCent/Action-Progression-Network