AI Paper Atlas

Research map / Video / Action recognition

D2ST-Adapter: Disentangled-and-Deformable Spatio-Temporal Adapter for Few-shot Action Recognition

Wenjie Pei, Qizhong Tan, Guangming Lu, Jiandong Tian, Jun Yu

ICCV 2025 · Video › Action recognition · approach: Spatiotemporal CNNs for action recognition

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
add 3D-conv spatio-temporal adapter instead of vanilla adapter (ablation)ResNet-50 few-shot · SSv2-Full 1-shotaccuracy (%): 45.3 -> 52.2Improved
disentangle spatial and temporal pathways (DST-Adapter) rather than joint ST-Adapter (ablation)ResNet-50 few-shot · SSv2-Full 1-shotaccuracy (%): 52.2 -> 53.7Improved
replace 3D convolution with anisotropic deformable attention (aDSTA) (ablation)DST-Adapter · SSv2-Full 1-shotaccuracy (%): 53.7 -> 57.0Improved
replace 3D convolution with aDSTA (ablation)DST-Adapter · Kinetics 1-shotaccuracy (%): 74.4 -> 75.8Improved
use both spatial and temporal aDSTA branches rather than spatial only (ablation)D2ST-Adapter · SSv2-Full 1-shotaccuracy (%): 55.2 -> 57.0Improved
use anisotropic sampling densities rather than uniform (ablation)D2ST-Adapter · SSv2-Full 1-shotaccuracy (%): 55.8 -> 57.0Improved
insert adapters in all 12 stages rather than only the last 6 (ablation)D2ST-Adapter CLIP-ViT-B · SSv2-Small 1-shotaccuracy (%): 54.2 -> 55.0Improved
insert adapters in last 6 stages instead of first 6 (ablation)D2ST-Adapter CLIP-ViT-B · SSv2-Small 1-shotaccuracy (%): 48.4 -> 54.2Improved
increase bottleneck ratio from 0.0625 to 0.25 (ablation)D2ST-Adapter CLIP-ViT-B · SSv2-Small 1-shotaccuracy (%): 53.4 -> 55.0Improved
increase bottleneck ratio from 0.25 to 0.5 (ablation)D2ST-Adapter CLIP-ViT-B · SSv2-Small 1-shotaccuracy (%): 55.0 -> 54.7No clear effect
increase sampling density from <1,4,4>&<4,2,2> to <2,4,4>&<8,2,2> (ablation)D2ST-Adapter CLIP-ViT-B · SSv2-Full 1-shotaccuracy (%): 65.9 -> 66.7Improved
increase input frames from 4 to 16 (ablation)D2ST-Adapter CLIP-ViT-B · SSv2-Full 1-shotaccuracy (%): 63.6 -> 69.9Improved

Later papers that compared with this method

2 later accepted papers report D2ST-Adapter as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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