Temporal Segment Networks: Towards Good Practices for Deep Action Recognition
arXiv:1608.00859
Abstract
Deep convolutional networks have achieved great success for visual recognition in still images. However, for action recognition in videos, the advantage over traditional methods is not so evident. This paper aims to discover the principles to design effective ConvNet architectures for action recognition in videos and learn these models given limited training samples. Our first contribution is temporal segment network (TSN), a novel framework for video-based action recognition. which is based on the idea of long-range temporal structure modeling. It combines a sparse temporal sampling strategy and video-level supervision to enable efficient and effective learning using the whole action video. The other contribution is our study on a series of good practices in learning ConvNets on video data with the help of temporal segment network. Our approach obtains the state-the-of-art performance on the datasets of HMDB51 ( ) and UCF101 (). We also visualize the learned ConvNet models, which qualitatively demonstrates the effectiveness of temporal segment network and the proposed good practices.
Accepted by ECCV 2016. Based on this method, we won the ActivityNet challenge 2016 in untrimmed video classification
References in corpus (3)
Cited by in corpus (69)
- Anabranch Network for Camouflaged Object Segmentation
- Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition
- Attention Bottlenecks for Multimodal Fusion
- SCAN: Self-and-Collaborative Attention Network for Video Person Re-identification
- BSN: Boundary Sensitive Network for Temporal Action Proposal Generation
- PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences
- CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos
- Revisiting the Effectiveness of Off-the-shelf Temporal Modeling Approaches for Large-scale Video Classification
- Keeping Your Eye on the Ball: Trajectory Attention in Video Transformers
- V4D:4D Convolutional Neural Networks for Video-level Representation Learning
- VideoGraph: Recognizing Minutes-Long Human Activities in Videos
- Skeleton-based Action Recognition of People Handling Objects
- Temporal Generative Adversarial Nets with Singular Value Clipping
- Multi-Modality Fusion based on Consensus-Voting and 3D Convolution for Isolated Gesture Recognition
- Self-Supervised Video Representation Learning with Space-Time Cubic Puzzles
- CT-Net: Channel Tensorization Network for Video Classification
- StNet: Local and Global Spatial-Temporal Modeling for Action Recognition
- Temporal Pyramid Network for Action Recognition
- A Deep Network for Arousal-Valence Emotion Prediction with Acoustic-Visual Cues
- Exploiting Spatial-Temporal Modelling and Multi-Modal Fusion for Human Action Recognition
- Taming Visually Guided Sound Generation
- Baidu-UTS Submission to the EPIC-Kitchens Action Recognition Challenge 2019
- Attention Clusters: Purely Attention Based Local Feature Integration for Video Classification
- PredNet and Predictive Coding: A Critical Review
- GTA: Global Temporal Attention for Video Action Understanding
- Multi-Fiber Networks for Video Recognition
- End-to-end Video-level Representation Learning for Action Recognition
- Temporal Interlacing Network
- Long Short-Term Transformer for Online Action Detection
- Zero-Shot Visual Recognition via Bidirectional Latent Embedding
- Collaborative Distillation in the Parameter and Spectrum Domains for Video Action Recognition
- Compressed Video Action Recognition with Refined Motion Vector
- Intra- and Inter-Action Understanding via Temporal Action Parsing
- Exploiting Temporal Relationships in Video Moment Localization with Natural Language
- Move Forward and Tell: A Progressive Generator of Video Descriptions
- MotionSqueeze: Neural Motion Feature Learning for Video Understanding
- Modality-Balanced Models for Visual Dialogue
- Pillar Networks++: Distributed non-parametric deep and wide networks
- Two-stream Convolutional Networks for Multi-frame Face Anti-spoofing
- Low-Fidelity End-to-End Video Encoder Pre-training for Temporal Action Localization
- LNSMM: Eye Gaze Estimation With Local Network Share Multiview Multitask
- Action Recognition with Coarse-to-Fine Deep Feature Integration and Asynchronous Fusion
- Distilling Audio-Visual Knowledge by Compositional Contrastive Learning
- Interpretable and Trustworthy Deepfake Detection via Dynamic Prototypes
- Generalized Zero-Shot Learning for Action Recognition with Web-Scale Video Data
- Learning to Compose Topic-Aware Mixture of Experts for Zero-Shot Video Captioning
- Scheduled Differentiable Architecture Search for Visual Recognition
- Motion Feature Network: Fixed Motion Filter for Action Recognition
- Recognizing Video Events with Varying Rhythms
- Video Representation Learning and Latent Concept Mining for Large-scale Multi-label Video Classification
- Temporal Bilinear Networks for Video Action Recognition
- LIGAR: Lightweight General-purpose Action Recognition
- Symbiotic Attention with Privileged Information for Egocentric Action Recognition
- Video Question Generation via Cross-Modal Self-Attention Networks Learning
- Knowledge Integration Networks for Action Recognition
- Cricket stroke extraction: Towards creation of a large-scale cricket actions dataset
- Learning Spatio-Temporal Representation with Local and Global Diffusion
- Learning Spatiotemporal Features for Infrared Action Recognition with 3D Convolutional Neural Networks
- TimeGate: Conditional Gating of Segments in Long-range Activities
- Multi-modal Residual Perceptron Network for Audio-Video Emotion Recognition
- Video Contrastive Learning with Global Context
- Towards Good Practices for Multi-modal Fusion in Large-scale Video Classification
- Human Activity Recognition for Edge Devices
- Semantic Adversarial Network with Multi-scale Pyramid Attention for Video Classification
- T-RECS: Training for Rate-Invariant Embeddings by Controlling Speed for Action Recognition
- Large-Scale Mapping of Human Activity using Geo-Tagged Videos
- Approximated Bilinear Modules for Temporal Modeling
- Nuisance-Label Supervision: Robustness Improvement by Free Labels
- Depth Guided Adaptive Meta-Fusion Network for Few-shot Video Recognition