AI Paper Atlas

Research map / Video / Action recognition

VideoMamba: Spatio-Temporal Selective State Space Model

Jinyoung Park, Hee-Seon Kim, Kangwook Ko, Minbeom Kim, Changick Kim

ECCV 2024 · Video › Action recognition · approach: Contrastive video representation learning · 46 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
pretrain on ImageNet-1K instead of training from scratch (ablation)VideoMamba · HMDB51Top-1 accuracy: 18.1 -> 59.3Improved
pretrain on Kinetics-400 instead of ImageNet-1K (ablation)VideoMamba · HMDB51Top-1 accuracy: 59.3 -> 68.6Improved
pretrain on Kinetics-400 instead of ImageNet-1K (ablation)VideoMamba · Something-Something V2Top-1 accuracy: 63.7 -> 63.9No clear effect
increase input frames from 16 to 32 (ablation)VideoMamba · Kinetics-400 valTop-1 accuracy: 76.1 -> 77.7Improved
increase input frames from 16 to 32 with K400 pretraining (ablation)VideoMamba · HMDB51Top-1 accuracy: 68.6 -> 75.7Improved
use spatio-temporal reversal for backward SSM instead of spatial reversal (ablation)VideoMamba · SSV2 and HMDB51Top-1 accuracyImproved
add temporal expansion initialization of learnable positional embeddings (ablation)VideoMamba · SSV2 and HMDB51Top-1 accuracyImproved
add RandAugment and label smoothing (ablation)VideoMamba · HMDB51Top-1 accuracy: about +1% eachImproved
reorder input frames progressively more severely (blockwise, pairwise, interleaved) (ablation)VideoMamba · HMDB51Top-1 accuracyHurt
replace VideoSwin-T with VideoMamba on SSV2VideoMamba · Something-Something V2 valTop-1 accuracy: 57.2 -> 63.7Improved
replace VideoSwin-T with VideoMamba on Kinetics-400VideoMamba · Kinetics-400 valTop-1 accuracy: 78.8 -> 77.7Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy