# Deep Local Video Feature for Action Recognition

arXiv: https://arxiv.org/abs/1701.07368 · HTML: https://mlsanity.com/paper/1701.07368/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 25 Jan 2017.

## Built on by (most cited first, 10 of 10)

- [Video Contrastive Learning with Global Context](https://mlsanity.com/paper/2108.02722/) (2021)
- [Pose-guided Inter- and Intra-part Relational Transformer for Occluded Person Re-Identification](https://mlsanity.com/paper/2109.03483/) (2021)
- [Deep Two-Stream Video Inference for Human Body Pose and Shape Estimation](https://mlsanity.com/paper/2110.11680/) (2021)
- [Interpretable Deep Feature Propagation for Early Action Recognition](https://mlsanity.com/paper/2107.05122/) (2021)
- [DynImg: Key Frames with Visual Prompts are Good Representation for Multi-Modal Video Understanding](https://mlsanity.com/paper/2507.15569/) (2025)
- [A Survey on Backbones for Deep Video Action Recognition](https://mlsanity.com/paper/2405.05584/) (2024)
- [Three-stream network for enriched Action Recognition](https://mlsanity.com/paper/2104.13051/) (2021)
- [Cross-Block Fine-Grained Semantic Cascade for Skeleton-Based Sports Action Recognition](https://mlsanity.com/paper/2404.19383/) (2024)
- [Searching Similarity Measure for Binarized Neural Networks](https://mlsanity.com/paper/2206.03325/) (2022)
- [Texture-Based Input Feature Selection for Action Recognition](https://mlsanity.com/paper/2303.00138/) (2023)

## Abstract (arXiv)

We investigate the problem of representing an entire video using CNN features for human action recognition. Currently, limited by GPU memory, we have not been able to feed a whole video into CNN/RNNs for end-to-end learning. A common practice is to use sampled frames as inputs and video labels as supervision. One major problem of this popular approach is that the local samples may not contain the information indicated by global labels. To deal with this problem, we propose to treat the deep networks trained on local inputs as local feature extractors. After extracting local features, we aggregate them into global features and train another mapping function on the same training data to map the global features into global labels. We study a set of problems regarding this new type of local features such as how to aggregate them into global features. Experimental results on HMDB51 and UCF101 datasets show that, for these new local features, a simple maximum pooling on the sparsely sampled features lead to significant performance improvement.