Spatio-temporal Relation Modeling for Few-shot Action Recognition | CVPR 2022

Paper Abstract:
Novel Few-shot action recognition framework, STRM, is proposed for learning higher-order temporal representations. Aggregate spatial and temporal contexts with dedicated local patch-level and global frame-level feature enrichment sub-modules. Propose a query-class similarity classifier on the patch-level enriched features to enhance class-specific feature discriminability by reinforcing the feature learning at different stages in the proposed framework. Achieved an absolute gain of 3.5 % in classification accuracy over the existing methods on the challenging SSv2 benchmark.
Speaker Bio: Incoming MS CS student at University of California, Riverside.
B.Tech. in CSE, Shiv Nadar University,
Ex-research Assistant at IIIT-Hyderabad,
Ex-research Intern at MBZUAI.
Link to Presentation: docs.google.com/presentation/...
Personal links:
/ athatipelli
github.com/Anirudh257

[@rrverma6391]

thanks