Training deep learning based video classifiers for action recognition requires a large amount of labeled videos. The labeling process is labor-intensive and time-consuming. On the other hand, large amount of weakly-labeled images are uploaded to the Internet by users everyday. To harness the rich and highly diverse set of Web images, a scalable approach is to crawl these images to train deep learning based classifier, such as Convolutional Neural Networks (CNN). However, due to the domain shift problem, the performance of Web images trained deep classifiers tend to degrade when directly deployed to videos. One way to address this problem is to fine-tune the trained models on videos, but sufficient amount of annotated videos are still required. In this work, we propose a novel approach to transfer knowledge from image domain to video domain. The proposed method can adapt to the target domain (i.e. video data) with limited amount of training data. Our method maps the video frames into a low-dimensional feature space using the class-discriminative spatial attention map for CNNs. We design a novel Siamese EnergyNet structure to learn energy functions on the attention maps by jointly optimizing two loss functions, such that the attention map corresponding to a ground truth concept would have higher energy. We conduct extensive experiments on two challenging video recognition datasets (i.e. TVHI and UCF101), and demonstrate the efficacy of our proposed method.
Current state-of-the-art human action recognition is focused on the classification of temporally trimmed videos in which only one action occurs per frame. In this work we address the problem of action localisation and instance segmentation in which multiple concurrent actions of the same class may be segmented out of an image sequence. We cast the action tube extraction as an energy maximisation problem in which configurations of region proposals in each frame are assigned a cost and the best action tubes are selected via two passes of dynamic programming. One pass associates region proposals in space and time for each action category, and another pass is used to solve for the tube's temporal extent and to enforce a smooth label sequence through the video. In addition, by taking advantage of recent work on action foreground-background segmentation, we are able to associate each tube with class-specific segmentations. We demonstrate the performance of our algorithm on the challenging LIRIS-HARL dataset and achieve a new state-of-the-art result which is 14.3 times better than previous methods.
Skeleton-based human action recognition has recently attracted increasing attention due to the popularity of 3D skeleton data. One main challenge lies in the large view variations in captured human actions. We propose a novel view adaptation scheme to automatically regulate observation viewpoints during the occurrence of an action. Rather than re-positioning the skeletons based on a human defined prior criterion, we design a view adaptive recurrent neural network (RNN) with LSTM architecture, which enables the network itself to adapt to the most suitable observation viewpoints from end to end. Extensive experiment analyses show that the proposed view adaptive RNN model strives to (1) transform the skeletons of various views to much more consistent viewpoints and (2) maintain the continuity of the action rather than transforming every frame to the same position with the same body orientation. Our model achieves significant improvement over the state-of-the-art approaches on three benchmark datasets.
Temporal Action Proposal (TAP) generation is an important problem, as fast and accurate extraction of semantically important (e.g. human actions) segments from untrimmed videos is an important step for large-scale video analysis. We propose a novel Temporal Unit Regression Network (TURN) model. There are two salient aspects of TURN: (1) TURN jointly predicts action proposals and refines the temporal boundaries by temporal coordinate regression; (2) Fast computation is enabled by unit feature reuse: a long untrimmed video is decomposed into video units, which are reused as basic building blocks of temporal proposals. TURN outperforms the state-of-the-art methods under average recall (AR) by a large margin on THUMOS-14 and ActivityNet datasets, and runs at over 880 frames per second (FPS) on a TITAN X GPU. We further apply TURN as a proposal generation stage for existing temporal action localization pipelines, it outperforms state-of-the-art performance on THUMOS-14 and ActivityNet.
We address human action recognition from multi-modal video data involving articulated pose and RGB frames and propose a two-stream approach. The pose stream is processed with a convolutional model taking as input a 3D tensor holding data from a sub-sequence. A specific joint ordering, which respects the topology of the human body, ensures that different convolutional layers correspond to meaningful levels of abstraction. The raw RGB stream is handled by a spatio-temporal soft-attention mechanism conditioned on features from the pose network. An LSTM network receives input from a set of image locations at each instant. A trainable glimpse sensor extracts features on a set of predefined locations specified by the pose stream, namely the 4 hands of the two people involved in the activity. Appearance features give important cues on hand motion and on objects held in each hand. We show that it is of high interest to shift the attention to different hands at different time steps depending on the activity itself. Finally a temporal attention mechanism learns how to fuse LSTM features over time. We evaluate the method on 3 datasets. State-of-the-art results are achieved on the largest dataset for human activity recognition, namely NTU-RGB+D, as well as on the SBU Kinect Interaction dataset. Performance close to state-of-the-art is achieved on the smaller MSR Daily Activity 3D dataset.
Existing RNN-based approaches for action recognition from depth sequences require either skeleton joints or hand-crafted depth features as inputs. An end-to-end manner, mapping from raw depth maps to action classes, is non-trivial to design due to the fact that: 1) single channel map lacks texture thus weakens the discriminative power; 2) relatively small set of depth training data. To address these challenges, we propose to learn an RNN driven by privileged information (PI) in three-steps: An encoder is pre-trained to learn a joint embedding of depth appearance and PI (i.e. skeleton joints). The learned embedding layers are then tuned in the learning step, aiming to optimize the network by exploiting PI in a form of multi-task loss. However, exploiting PI as a secondary task provides little help to improve the performance of a primary task (i.e. classification) due to the gap between them. Finally, a bridging matrix is defined to connect two tasks by discovering latent PI in the refining step. Our PI-based classification loss maintains a consistency between latent PI and predicted distribution. The latent PI and network are iteratively estimated and updated in an expectation-maximization procedure. The proposed learning process provides greater discriminative power to model subtle depth difference, while helping avoid overfitting the scarcer training data. Our experiments show significant performance gains over state-of-the-art methods on three public benchmark datasets and our newly collected Blanket dataset.
Human actions captured in video sequences are three-dimensional signals characterizing visual appearance and motion dynamics. To learn action patterns, existing methods adopt Convolutional and/or Recurrent Neural Networks (CNNs and RNNs). CNN based methods are effective in learning spatial appearances, but are limited in modeling long-term motion dynamics. RNNs, especially Long Short-Term Memory (LSTM), are able to learn temporal motion dynamics. However, naively applying RNNs to video sequences in a convolutional manner implicitly assumes that motions in videos are stationary across different spatial locations. This assumption is valid for short-term motions but invalid when the duration of the motion is long. In this work, we propose Lattice-LSTM (L2STM), which extends LSTM by learning independent hidden state transitions of memory cells for individual spatial locations. This method effectively enhances the ability to model dynamics across time and addresses the non-stationary issue of long-term motion dynamics without significantly increasing the model complexity. Additionally, we introduce a novel multi-modal training procedure for training our network. Unlike traditional two-stream architectures which use RGB and optical flow information as input, our two-stream model leverages both modalities to jointly train both input gates and both forget gates in the network rather than treating the two streams as separate entities with no information about the other. We apply this end-to-end system to benchmark datasets (UCF-101 and HMDB-51) of human action recognition. Experiments show that on both datasets, our proposed method outperforms all existing ones that are based on LSTM and/or CNNs of similar model complexities.
We study the problem of video classification for facial analysis and human action recognition. We propose a novel weakly supervised learning method that models the video as a sequence of automatically mined, discriminative sub-events (eg. onset and offset phase for "smile", running and jumping for "highjump"). The proposed model is inspired by the recent works on Multiple Instance Learning and latent SVM/HCRF -- it extends such frameworks to model the ordinal aspect in the videos, approximately. We obtain consistent improvements over relevant competitive baselines on four challenging and publicly available video based facial analysis datasets for prediction of expression, clinical pain and intent in dyadic conversations and on three challenging human action datasets. We also validate the method with qualitative results and show that they largely support the intuitions behind the method.
In recent work, it was shown that combining multi-kernel based support vector machines (SVMs) can lead to near state-of-the-art performance on an action recognition dataset (HMDB-51 dataset). This was 0.4\% lower than frameworks that used hand-crafted features in addition to the deep convolutional feature extractors. In the present work, we show that combining distributed Gaussian Processes with multi-stream deep convolutional neural networks (CNN) alleviate the need to augment a neural network with hand-crafted features. In contrast to prior work, we treat each deep neural convolutional network as an expert wherein the individual predictions (and their respective uncertainties) are combined into a Product of Experts (PoE) framework.
We introduce Eigen Evolution Pooling, an efficient method to aggregate a sequence of feature vectors. Eigen evolution pooling is designed to produce compact feature representations for a sequence of feature vectors, while maximally preserving as much information about the sequence as possible, especially the temporal evolution of the features over time. Eigen evolution pooling is a general pooling method that can be applied to any sequence of feature vectors, from low-level RGB values to high-level Convolutional Neural Network (CNN) feature vectors. We show that eigen evolution pooling is more effective than average, max, and rank pooling for encoding the dynamics of human actions in video. We demonstrate the power of eigen evolution pooling on UCF101 and Hollywood2 datasets, two human action recognition benchmarks, and achieve state-of-the-art performance.
Human action recognition has been one of the most active fields of research in computer vision for last years. Two dimensional action recognition methods are facing serious challenges such as occlusion and missing the third dimension of data. Development of depth sensors has made it feasible to track positions of human body joints over time. This paper proposes a novel method of action recognition which uses temporal 3D skeletal Kinect data. This method introduces the definition of body states and then every action is modeled as a sequence of these states. The learning stage uses Fisher Linear Discriminant Analysis (LDA) to construct discriminant feature space for discriminating the body states. Moreover, this paper suggests the use of the Mahalonobis distance as an appropriate distance metric for the classification of the states of involuntary actions. Hidden Markov Model (HMM) is then used to model the temporal transition between the body states in each action. According to the results, this method significantly outperforms other popular methods, with recognition rate of 88.64% for eight different actions and up to 96.18% for classifying fall actions.
Current state-of-the-art approaches for spatio-temporal action localization rely on detections at the frame level that are then linked or tracked across time. In this paper, we leverage the temporal continuity of videos instead of operating at the frame level. We propose the ACtion Tubelet detector (ACT-detector) that takes as input a sequence of frames and outputs tubelets, i.e., sequences of bounding boxes with associated scores. The same way state-of-the-art object detectors rely on anchor boxes, our ACT-detector is based on anchor cuboids. We build upon the SSD framework. Convolutional features are extracted for each frame, while scores and regressions are based on the temporal stacking of these features, thus exploiting information from a sequence. Our experimental results show that leveraging sequences of frames significantly improves detection performance over using individual frames. The gain of our tubelet detector can be explained by both more accurate scores and more precise localization. Our ACT-detector outperforms the state-of-the-art methods for frame-mAP and video-mAP on the J-HMDB and UCF-101 datasets, in particular at high overlap thresholds.
We introduce the concept of "dynamic image", a novel compact representation of videos useful for video analysis, particularly in combination with convolutional neural networks (CNNs). A dynamic image encodes temporal data such as RGB or optical flow videos by using the concept of `rank pooling'. The idea is to learn a ranking machine that captures the temporal evolution of the data and to use the parameters of the latter as a representation. When a linear ranking machine is used, the resulting representation is in the form of an image, which we call dynamic because it summarizes the video dynamics in addition of appearance. This is a powerful idea because it allows to convert any video to an image so that existing CNN models pre-trained for the analysis of still images can be immediately extended to videos. We also present an efficient and effective approximate rank pooling operator, accelerating standard rank pooling algorithms by orders of magnitude, and formulate that as a CNN layer. This new layer allows generalizing dynamic images to dynamic feature maps. We demonstrate the power of the new representations on standard benchmarks in action recognition achieving state-of-the-art performance.
Human action recognition involves the characterization of human actions through the automated analysis of video data and is integral in the development of smart computer vision systems. However, several challenges like dynamic backgrounds, camera stabilization, complex actions, occlusions etc. make action recognition in a real time and robust fashion difficult. Several complex approaches exist but are computationally intensive. This paper presents a novel approach of using a combination of good features along with iterative optical flow algorithm to compute feature vectors which are classified using a multilayer perceptron (MLP) network. The use of multiple features for motion descriptors enhances the quality of tracking. Resilient backpropagation algorithm is used for training the feedforward neural network reducing the learning time. The overall system accuracy is improved by optimizing the various parameters of the multilayer perceptron network.
Convolutional neural networks with spatio-temporal 3D kernels (3D CNNs) have an ability to directly extract spatio-temporal features from videos for action recognition. Although the 3D kernels tend to overfit because of a large number of their parameters, the 3D CNNs are greatly improved by using recent huge video databases. However, the architecture of 3D CNNs is relatively shallow against to the success of very deep neural networks in 2D-based CNNs, such as residual networks (ResNets). In this paper, we propose a 3D CNNs based on ResNets toward a better action representation. We describe the training procedure of our 3D ResNets in details. We experimentally evaluate the 3D ResNets on the ActivityNet and Kinetics datasets. The 3D ResNets trained on the Kinetics did not suffer from overfitting despite the large number of parameters of the model, and achieved better performance than relatively shallow networks, such as C3D. Our code and pretrained models (e.g. Kinetics and ActivityNet) are publicly available at https://github.com/kenshohara/3D-ResNets.
Recurrent Neural Networks (RNNs) have been widely used in natural language processing and computer vision. Among them, the Hierarchical Multi-scale RNN (HM-RNN), a kind of multi-scale hierarchical RNN proposed recently, can learn the hierarchical temporal structure from data automatically. In this paper, we extend the work to solve the computer vision task of action recognition. However, in sequence-to-sequence models like RNN, it is normally very hard to discover the relationships between inputs and outputs given static inputs. As a solution, attention mechanism could be applied to extract the relevant information from input thus facilitating the modeling of input-output relationships. Based on these considerations, we propose a novel attention network, namely Hierarchical Multi-scale Attention Network (HM-AN), by combining the HM-RNN and the attention mechanism and apply it to action recognition. A newly proposed gradient estimation method for stochastic neurons, namely Gumbel-softmax, is exploited to implement the temporal boundary detectors and the stochastic hard attention mechanism. To amealiate the negative effect of sensitive temperature of the Gumbel-softmax, an adaptive temperature training method is applied to better the system performance. The experimental results demonstrate the improved effect of HM-AN over LSTM with attention on the vision task. Through visualization of what have been learnt by the networks, it can be observed that both the attention regions of images and the hierarchical temporal structure can be captured by HM-AN.
Many human activities involve object manipulations aiming to modify the object state. Examples of common state changes include full/empty bottle, open/closed door, and attached/detached car wheel. In this work, we seek to automatically discover the states of objects and the associated manipulation actions. Given a set of videos for a particular task, we propose a joint model that learns to identify object states and to localize state-modifying actions. Our model is formulated as a discriminative clustering cost with constraints. We assume a consistent temporal order for the changes in object states and manipulation actions, and introduce new optimization techniques to learn model parameters without additional supervision. We demonstrate successful discovery of seven manipulation actions and corresponding object states on a new dataset of videos depicting real-life object manipulations. We show that our joint formulation results in an improvement of object state discovery by action recognition and vice versa.
This paper proposes a two-stream flow-guided convolutional attention networks for action recognition in videos. The central idea is that optical flows, when properly compensated for the camera motion, can be used to guide attention to the human foreground. We thus develop cross-link layers from the temporal network (trained on flows) to the spatial network (trained on RGB frames). These cross-link layers guide the spatial-stream to pay more attention to the human foreground areas and be less affected by background clutter. We obtain promising performances with our approach on the UCF101, HMDB51 and Hollywood2 datasets.
This paper proposes a multi-level feature learning framework for human action recognition using a single body-worn inertial sensor. The framework consists of three phases, respectively designed to analyze signal-based (low-level), components (mid-level) and semantic (high-level) information. Low-level features capture the time and frequency domain property while mid-level representations learn the composition of the action. The Max-margin Latent Pattern Learning (MLPL) method is proposed to learn high-level semantic descriptions of latent action patterns as the output of our framework. The proposed method achieves the state-of-the-art performances, 88.7%, 98.8% and 72.6% (weighted F1 score) respectively, on Skoda, WISDM and OPP datasets.
Recently, there has been a growing interest in analyzing human daily activities from data collected by wearable cameras. Since the hands are involved in a vast set of daily tasks, detecting hands in egocentric images is an important step towards the recognition of a variety of egocentric actions. However, besides extreme illumination changes in egocentric images, hand detection is not a trivial task because of the intrinsic large variability of hand appearance. We propose a hand detector that exploits skin modeling for fast hand proposal generation and Convolutional Neural Networks for hand recognition. We tested our method on UNIGE-HANDS dataset and we showed that the proposed approach achieves competitive hand detection results.