Reproduce “Learning Representative Temporal Features for Action Recognition” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ali Javidani, Ahmad Mahmoudi-Aznaveh
The method recognizes video actions by converting optical flow images into feature sequences with fixed pretrained image networks and training a small one-dimensional convolutional neural network over time. Temporal pyramid pooling reaches 74.8% mean accuracy on jHMDB across three splits versus 72.3% with zero padding. Its 1D-CNN has 50K trainable parameters versus C3D’s 79M; this compares parameter counts, not runtime.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.