AI Paper Atlas

Research map / Video / Action recognition

Action Recognition with Dynamic Image Networks

Hakan Bilen, Basura Fernando, Efstratios Gavves, Andrea Vedaldi

TPAMI 2017 · Video › Action recognition · approach: Spatiotemporal CNNs for action recognition · 269 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
replace mean or max pooling with rank pooling for the single-image video summary (ablation)single dynamic image · UCF101 split-1accuracy (%): 52.6 -> 57.2 (mean), 48.0 -> 57.2 (max)Improved
use dynamic image instead of motion history image (ablation)single image CaffeNet · UCF101 split-1accuracy (%): 46.6 -> 57.2Improved
use approximate rank pooling instead of exact rank poolingdynamic image · UCF101 split-1fps: 131 -> 5920Improved
use approximate rank pooling instead of exact rank poolingdynamic image · UCF101 split-1classification accuracy (%): 57.9 -> 55.2Hurt
use approximate rank pooling instead of exact rank poolingdynamic image · UCF101 split-1pairwise ranking accuracy (%): 95.5 -> 96.5No clear effect
split video into multiple dynamic images (window 10, stride 6) instead of one per video (ablation)MDI · UCF101 split-1accuracy (%): 55.2 -> 67.4Improved
increase window size to 20 with stride 12 (ablation)MDI · UCF101 split-1accuracy (%): 67.4 -> 58.9Hurt
use max pooling over dynamic images instead of mean or ARP in temporal pooling layer (ablation)MDI · UCF101 split-1accuracy (%): 66.2 -> 68.3, 65.2 -> 68.3Improved
move ARP from input (conv1) to before conv3 (ablation)dynamic map network · UCF101accuracy (%): 70.9 -> 73.3Improved
place ARP before conv4 instead of conv3 (ablation)dynamic map network · UCF101accuracy (%): 73.3 -> 67.6Hurt
learn pooling coefficients (parametric pooling) instead of fixed ARP (ablation)dynamic image network · UCF101 and HMDB51accuracy (%): ARP conv1 70.9 vs PP1 67.6; ARP conv2 68.5 vs PP2 73.9No clear effect
switch CaffeNet to ResNeXt-50SI and DI streams · UCF101accuracy (%): 70.9 -> 86.6 (DI)Improved

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