Reproduce “Adaptive Residual Quantization for Memory-Efficient Temporal Action Segmentation” 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
Gerard Donahue, Guven Gergerli, Ayush Gupta, Reza Ghoddoosian, Enna Sachdeva, Faizan Siddiqui
The authors introduce ARQ-VAE, which compresses pre-extracted training features for temporal action segmentation using residual codes and frame-specific code lengths selected by reconstruction error. On Breakfast with ASFormer, fine-tuned ARQ-VAE* stores 10 MB and scores 71.5 Edit and 51.2 F1@50, versus GNI’s 44 MB, 48.8 Edit, and 34.1 F1@50; reported storage excludes the decoder.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.