Recently, few-shot learning has received increasing interest. Existing efforts have been focused on image classification, with very few attempts dedicated to the more challenging few-shot video classification problem. These few attempts aim to effectively exploit the temporal dimension in videos for better learning in low data regimes. However, they have largely ignored a key characteristic of video which could be vital for few-shot recognition, that is, videos are often accompanied by rich text descriptions. In this paper, for the first time, we propose to leverage these human-provided textual descriptions as privileged information when training a few-shot video classification model. Specifically, we formulate a text-based task conditioner to adapt video features to the few-shot learning task. Our model follows a transductive setting where query samples and support textual descriptions can be used to update the support set class prototype to further improve the task-adaptation ability of the model. Our model obtains state-of-the-art performance on four challenging benchmarks in few-shot video action classification.
Instance-level contrastive learning techniques, which rely on data augmentation and a contrastive loss function, have found great success in the domain of visual representation learning. They are not suitable for exploiting the rich dynamical structure of video however, as operations are done on many augmented instances. In this paper we propose "Video Cross-Stream Prototypical Contrasting", a novel method which predicts consistent prototype assignments from both RGB and optical flow views, operating on sets of samples. Specifically, we alternate the optimization process; while optimizing one of the streams, all views are mapped to one set of stream prototype vectors. Each of the assignments is predicted with all views except the one matching the prediction, pushing representations closer to their assigned prototypes. As a result, more efficient video embeddings with ingrained motion information are learned, without the explicit need for optical flow computation during inference. We obtain state-of-the-art results on nearest-neighbour video retrieval and action recognition, outperforming previous best by +3.2% on UCF101 using the S3D backbone (90.5% Top-1 acc), and by +7.2% on UCF101 and +15.1% on HMDB51 using the R(2+1)D backbone.
3D convolution is powerful for video classification but often computationally expensive, recent studies mainly focus on decomposing it on spatial-temporal and/or channel dimensions. Unfortunately, most approaches fail to achieve a preferable balance between convolutional efficiency and feature-interaction sufficiency. For this reason, we propose a concise and novel Channel Tensorization Network (CT-Net), by treating the channel dimension of input feature as a multiplication of K sub-dimensions. On one hand, it naturally factorizes convolution in a multiple dimension way, leading to a light computation burden. On the other hand, it can effectively enhance feature interaction from different channels, and progressively enlarge the 3D receptive field of such interaction to boost classification accuracy. Furthermore, we equip our CT-Module with a Tensor Excitation (TE) mechanism. It can learn to exploit spatial, temporal and channel attention in a high-dimensional manner, to improve the cooperative power of all the feature dimensions in our CT-Module. Finally, we flexibly adapt ResNet as our CT-Net. Extensive experiments are conducted on several challenging video benchmarks, e.g., Kinetics-400, Something-Something V1 and V2. Our CT-Net outperforms a number of recent SOTA approaches, in terms of accuracy and/or efficiency. The codes and models will be available on https://github.com/Andy1621/CT-Net.
Temporal action segmentation is a task to classify each frame in the video with an action label. However, it is quite expensive to annotate every frame in a large corpus of videos to construct a comprehensive supervised training dataset. Thus in this work we propose an unsupervised method, namely SSCAP, that operates on a corpus of unlabeled videos and predicts a likely set of temporal segments across the videos. SSCAP leverages Self-Supervised learning to extract distinguishable features and then applies a novel Co-occurrence Action Parsing algorithm to not only capture the correlation among sub-actions underlying the structure of activities, but also estimate the temporal path of the sub-actions in an accurate and general way. We evaluate on both classic datasets (Breakfast, 50Salads) and the emerging fine-grained action dataset (FineGym) with more complex activity structures and similar sub-actions. Results show that SSCAP achieves state-of-the-art performance on all datasets and can even outperform some weakly-supervised approaches, demonstrating its effectiveness and generalizability.
There has been an increasing interest in spiking neural networks in recent years. SNNs are seen as hypothetical solutions for the bottlenecks of ANNs in pattern recognition, such as energy efficiency. But current methods such as ANN-to-SNN conversion and back-propagation do not take full advantage of these networks, and unsupervised methods have not yet reached a success comparable to advanced artificial neural networks. It is important to study the behavior of SNNs trained with unsupervised learning methods such as spike-timing dependent plasticity (STDP) on video classification tasks, including mechanisms to model motion information using spikes, as this information is critical for video understanding. This paper presents multiple methods of transposing temporal information into a static format, and then transforming the visual information into spikes using latency coding. These methods are paired with two types of temporal fusion known as early and late fusion, and are used to help the spiking neural network in capturing the spatio-temporal features from videos. In this paper, we rely on the network architecture of a convolutional spiking neural network trained with STDP, and we test the performance of this network when challenged with action recognition tasks. Understanding how a spiking neural network responds to different methods of movement extraction and representation can help reduce the performance gap between SNNs and ANNs. In this paper we show the effect of the similarity in the shape and speed of certain actions on action recognition with spiking neural networks, we also highlight the effectiveness of some methods compared to others.
The absence or abnormality of fidgety movements of joints or limbs is strongly indicative of cerebral palsy in infants. Developing computer-based methods for assessing infant movements in videos is pivotal for improved cerebral palsy screening. Most existing methods use appearance-based features and are thus sensitive to strong but irrelevant signals caused by background clutter or a moving camera. Moreover, these features are computed over the whole frame, thus they measure gross whole body movements rather than specific joint/limb motion. Addressing these challenges, we develop and validate a new method for fidgety movement assessment from consumer-grade videos using human poses extracted from short clips. Human poses capture only relevant motion profiles of joints and limbs and are thus free from irrelevant appearance artifacts. The dynamics and coordination between joints are modeled using spatio-temporal graph convolutional networks. Frames and body parts that contain discriminative information about fidgety movements are selected through a spatio-temporal attention mechanism. We validate the proposed model on the cerebral palsy screening task using a real-life consumer-grade video dataset collected at an Australian hospital through the Cerebral Palsy Alliance, Australia. Our experiments show that the proposed method achieves the ROC-AUC score of 81.87%, significantly outperforming existing competing methods with better interpretability.
Understanding pedestrian crossing behavior is an essential goal in intelligent vehicle development, leading to an improvement in their security and traffic flow. In this paper, we developed a method called IntFormer. It is based on transformer architecture and a novel convolutional video classification model called RubiksNet. Following the evaluation procedure in a recent benchmark, we show that our model reaches state-of-the-art results with good performance ($\approx 40$ seq. per second) and size ($8\times $smaller than the best performing model), making it suitable for real-time usage. We also explore each of the input features, finding that ego-vehicle speed is the most important variable, possibly due to the similarity in crossing cases in PIE dataset.
Existing research on action recognition treats activities as monolithic events occurring in videos. Recently, the benefits of formulating actions as a combination of atomic-actions have shown promise in improving action understanding with the emergence of datasets containing such annotations, allowing us to learn representations capturing this information. However, there remains a lack of studies that extend action composition and leverage multiple viewpoints and multiple modalities of data for representation learning. To promote research in this direction, we introduce Home Action Genome (HOMAGE): a multi-view action dataset with multiple modalities and view-points supplemented with hierarchical activity and atomic action labels together with dense scene composition labels. Leveraging rich multi-modal and multi-view settings, we propose Cooperative Compositional Action Understanding (CCAU), a cooperative learning framework for hierarchical action recognition that is aware of compositional action elements. CCAU shows consistent performance improvements across all modalities. Furthermore, we demonstrate the utility of co-learning compositions in few-shot action recognition by achieving 28.6% mAP with just a single sample.
Few-shot video classification aims to learn new video categories with only a few labeled examples, alleviating the burden of costly annotation in real-world applications. However, it is particularly challenging to learn a class-invariant spatial-temporal representation in such a setting. To address this, we propose a novel matching-based few-shot learning strategy for video sequences in this work. Our main idea is to introduce an implicit temporal alignment for a video pair, capable of estimating the similarity between them in an accurate and robust manner. Moreover, we design an effective context encoding module to incorporate spatial and feature channel context, resulting in better modeling of intra-class variations. To train our model, we develop a multi-task loss for learning video matching, leading to video features with better generalization. Extensive experimental results on two challenging benchmarks, show that our method outperforms the prior arts with a sizable margin on SomethingSomething-V2 and competitive results on Kinetics.
We introduce Video Transformer (VidTr) with separable-attention for video classification. Comparing with commonly used 3D networks, VidTr is able to aggregate spatio-temporal information via stacked attentions and provide better performance with higher efficiency. We first introduce the vanilla video transformer and show that transformer module is able to perform spatio-temporal modeling from raw pixels, but with heavy memory usage. We then present VidTr which reduces the memory cost by 3.3$\times$ while keeping the same performance. To further optimize the model, we propose the standard deviation based topK pooling for attention ($pool_{topK\_std}$), which reduces the computation by dropping non-informative features along temporal dimension. VidTr achieves state-of-the-art performance on five commonly used datasets with lower computational requirement, showing both the efficiency and effectiveness of our design. Finally, error analysis and visualization show that VidTr is especially good at predicting actions that require long-term temporal reasoning.
Convolutional Neural Networks are extensively used in a wide range of applications, commonly including computer vision tasks like image and video classification, recognition, and segmentation. Recent research results demonstrate that multilayer(deep) networks involving mono-dimensional convolutions and dilation can be effectively used in time series and sequences classification and segmentation, as well as in tasks involving sequence modelling. These structures, commonly referred to as Temporal Convolutional Networks (TCNs), have been demonstrated to consistently outperform Recurrent Neural Networks in terms of accuracy and training time [1]. While FPGA-based inference accelerators for classic CNNs are widespread, literature is lacking in a quantitative evaluation of their usability on inference for TCN models. In this paper we present such an evaluation, considering a CNN accelerator with specific features supporting TCN kernels as a reference and a set of state-of-the-art TCNs as a benchmark. Experimental results show that, during TCN execution, operational intensity can be critical for the overall performance. We propose a convolution scheduling based on batch processing that can boost efficiency up to 96% of theoretical peak performance. Overall we can achieve up to 111,8 GOPS/s and power efficiency of 33,9 GOPS/s/W on an Ultrascale+ ZU3EG (up to 10x speedup and 3x power efficiency improvement with respect to pure software implementation).
Deep learning models need large amounts of data for training. In video recognition and classification, significant advances were achieved with the introduction of new large databases. However, the creation of large-databases for training is infeasible in several scenarios. Thus, existing or small collected databases are typically joined and amplified to train these models. Nevertheless, training neural networks on limited data is not straightforward and comes with a set of problems. In this paper, we explore the effects of stacking databases, model initialization, and data amplification techniques when training with limited data on deep learning models' performance. We focused on the problem of Facial Expression Recognition from videos. We performed an extensive study with four databases at a different complexity and nine deep-learning architectures for video classification. We found that (i) complex training sets translate better to more stable test sets when trained with transfer learning and synthetically generated data, but their performance yields a high variance; (ii) training with more detailed data translates to more stable performance on novel scenarios (albeit with lower performance); (iii) merging heterogeneous data is not a straightforward improvement, as the type of augmentation and initialization is crucial; (iv) classical data augmentation cannot fill the holes created by joining largely separated datasets; and (v) inductive biases help to bridge the gap when paired with synthetic data, but this data is not enough when working with standard initialization techniques.
The standard way of training video models entails sampling at each iteration a single clip from a video and optimizing the clip prediction with respect to the video-level label. We argue that a single clip may not have enough temporal coverage to exhibit the label to recognize, since video datasets are often weakly labeled with categorical information but without dense temporal annotations. Furthermore, optimizing the model over brief clips impedes its ability to learn long-term temporal dependencies. To overcome these limitations, we introduce a collaborative memory mechanism that encodes information across multiple sampled clips of a video at each training iteration. This enables the learning of long-range dependencies beyond a single clip. We explore different design choices for the collaborative memory to ease the optimization difficulties. Our proposed framework is end-to-end trainable and significantly improves the accuracy of video classification at a negligible computational overhead. Through extensive experiments, we demonstrate that our framework generalizes to different video architectures and tasks, outperforming the state of the art on both action recognition (e.g., Kinetics-400 & 700, Charades, Something-Something-V1) and action detection (e.g., AVA v2.1 & v2.2).
Video classification and analysis is always a popular and challenging field in computer vision. It is more than just simple image classification due to the correlation with respect to the semantic contents of subsequent frames brings difficulties for video analysis. In this literature review, we summarized some state-of-the-art methods for multi-label video classification. Our goal is first to experimentally research the current widely used architectures, and then to develop a method to deal with the sequential data of frames and perform multi-label classification based on automatic content detection of video.
In video data, busy motion details from moving regions are conveyed within a specific frequency bandwidth in the frequency domain. Meanwhile, the rest of the frequencies of video data are encoded with quiet information with substantial redundancy, which causes low processing efficiency in existing video models that take as input raw RGB frames. In this paper, we consider allocating intenser computation for the processing of the important busy information and less computation for that of the quiet information. We design a trainable Motion Band-Pass Module (MBPM) for separating busy information from quiet information in raw video data. By embedding the MBPM into a two-pathway CNN architecture, we define a Busy-Quiet Net (BQN). The efficiency of BQN is determined by avoiding redundancy in the feature space processed by the two pathways: one operating on Quiet features of low-resolution, while the other processes Busy features. The proposed BQN outperforms many recent video processing models on Something-Something V1, Kinetics400, UCF101 and HMDB51 datasets.
We present pure-transformer based models for video classification, drawing upon the recent success of such models in image classification. Our model extracts spatio-temporal tokens from the input video, which are then encoded by a series of transformer layers. In order to handle the long sequences of tokens encountered in video, we propose several, efficient variants of our model which factorise the spatial- and temporal-dimensions of the input. Although transformer-based models are known to only be effective when large training datasets are available, we show how we can effectively regularise the model during training and leverage pretrained image models to be able to train on comparatively small datasets. We conduct thorough ablation studies, and achieve state-of-the-art results on multiple video classification benchmarks including Kinetics 400 and 600, Epic Kitchens, Something-Something v2 and Moments in Time, outperforming prior methods based on deep 3D convolutional networks. To facilitate further research, we release code at https://github.com/google-research/scenic/tree/main/scenic/projects/vivit
Novel computer vision architectures monopolize the spotlight, but the impact of the model architecture is often conflated with simultaneous changes to training methodology and scaling strategies. Our work revisits the canonical ResNet (He et al., 2015) and studies these three aspects in an effort to disentangle them. Perhaps surprisingly, we find that training and scaling strategies may matter more than architectural changes, and further, that the resulting ResNets match recent state-of-the-art models. We show that the best performing scaling strategy depends on the training regime and offer two new scaling strategies: (1) scale model depth in regimes where overfitting can occur (width scaling is preferable otherwise); (2) increase image resolution more slowly than previously recommended (Tan & Le, 2019). Using improved training and scaling strategies, we design a family of ResNet architectures, ResNet-RS, which are 1.7x - 2.7x faster than EfficientNets on TPUs, while achieving similar accuracies on ImageNet. In a large-scale semi-supervised learning setup, ResNet-RS achieves 86.2% top-1 ImageNet accuracy, while being 4.7x faster than EfficientNet NoisyStudent. The training techniques improve transfer performance on a suite of downstream tasks (rivaling state-of-the-art self-supervised algorithms) and extend to video classification on Kinetics-400. We recommend practitioners use these simple revised ResNets as baselines for future research.
Although 3D Convolutional Neural Networks are essential for most learning based applications involving dense 3D data, their applicability is limited due to excessive memory and computational requirements. Compressing such networks by pruning therefore becomes highly desirable. However, pruning 3D CNNs is largely unexplored possibly because of the complex nature of typical pruning algorithms that embeds pruning into an iterative optimization paradigm. In this work, we introduce a Resource Aware Neuron Pruning (RANP) algorithm that prunes 3D CNNs at initialization to high sparsity levels. Specifically, the core idea is to obtain an importance score for each neuron based on their sensitivity to the loss function. This neuron importance is then reweighted according to the neuron resource consumption related to FLOPs or memory. We demonstrate the effectiveness of our pruning method on 3D semantic segmentation with widely used 3D-UNets on ShapeNet and BraTS'18 datasets, video classification with MobileNetV2 and I3D on UCF101 dataset, and two-view stereo matching with Pyramid Stereo Matching (PSM) network on SceneFlow dataset. In these experiments, our RANP leads to roughly 50%-95% reduction in FLOPs and 35%-80% reduction in memory with negligible loss in accuracy compared to the unpruned networks. This significantly reduces the computational resources required to train 3D CNNs. The pruned network obtained by our algorithm can also be easily scaled up and transferred to another dataset for training.
Quantizing deep networks with adaptive bit-widths is a promising technique for efficient inference across many devices and resource constraints. In contrast to static methods that repeat the quantization process and train different models for different constraints, adaptive quantization enables us to flexibly adjust the bit-widths of a single deep network during inference for instant adaptation in different scenarios. While existing research shows encouraging results on common image classification benchmarks, this paper investigates how to train such adaptive networks more effectively. Specifically, we present two novel techniques for quantizing deep neural networks with adaptive bit-widths of weights and activations. First, we propose a collaborative strategy to choose a high-precision teacher for transferring knowledge to the low-precision student while jointly optimizing the model with all bit-widths. Second, to effectively transfer knowledge, we develop a dynamic block swapping method by randomly replacing the blocks in the lower-precision student network with the corresponding blocks in the higher-precision teacher network. Extensive experiments on multiple image classification datasets including video classification benchmarks for the first time, well demonstrate the efficacy of our approach over state-of-the-art methods.
Since their inception, learning techniques under the Reservoir Computing paradigm have shown a great modeling capability for recurrent systems without the computing overheads required for other approaches. Among them, different flavors of echo state networks have attracted many stares through time, mainly due to the simplicity and computational efficiency of their learning algorithm. However, these advantages do not compensate for the fact that echo state networks remain as black-box models whose decisions cannot be easily explained to the general audience. This work addresses this issue by conducting an explainability study of Echo State Networks when applied to learning tasks with time series, image and video data. Specifically, the study proposes three different techniques capable of eliciting understandable information about the knowledge grasped by these recurrent models, namely, potential memory, temporal patterns and pixel absence effect. Potential memory addresses questions related to the effect of the reservoir size in the capability of the model to store temporal information, whereas temporal patterns unveils the recurrent relationships captured by the model over time. Finally, pixel absence effect attempts at evaluating the effect of the absence of a given pixel when the echo state network model is used for image and video classification. We showcase the benefits of our proposed suite of techniques over three different domains of applicability: time series modeling, image and, for the first time in the related literature, video classification. Our results reveal that the proposed techniques not only allow for a informed understanding of the way these models work, but also serve as diagnostic tools capable of detecting issues inherited from data (e.g. presence of hidden bias).