AI Paper Atlas

Research map / Video / Action recognition

Privacy-Preserving Deep Action Recognition: An Adversarial Learning Framework and A New Dataset

Zhenyu Wu, Haotao Wang, Zhaowen Wang, Hailin Jin, Zhangyang Wang

TPAMI 2020 · Video › Action recognition · approach: Disentangled action recognition · 97 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 GRL with negative-entropy objective (Ours-Entropy M=1)adversarial anonymization · SBUprivacy budget accuracy A_B (%): 85.4 -> 73.6Improved
add model restarting to Ours-Entropy (M=1) (ablation)Ours-Entropy · SBUprivacy budget accuracy A_B (%): 73.6 -> 60.5Improved
increase ensemble size from M=1 to M=8 with restarting (ablation)Ours-Entropy · SBUprivacy budget accuracy A_B (%): 60.5 -> 47.7Improved
add model restarting to K-Beam (K=2) (ablation)Ours-K-Beam · SBUprivacy budget accuracy A_B (%): 87.7 -> 80.6Improved
add model restarting to GRL (ablation)GRL · SBUprivacy budget accuracy A_B (%): 85.4 -> 84.3Improved
add model restarting to Ours-Entropy (M=4) on SBU action accuracy (ablation)Ours-Entropy · SBUaction accuracy A_T (%): 82.7 -> 78.0Hurt
increase K-Beam beams from K=4 to K=8 (ablation)Ours-K-Beam · SBUprivacy budget accuracy A_B (%): 88.4 -> 80.2Mixed
use naive downsampling rate 16 instead of 1Downsample · SBUaction accuracy A_T (%): 88.8 -> 64.4Hurt
apply Ours-Entropy with ensemble and restartingOurs-Entropy · UCF101/VISPRprivacy budget accuracy A_B (%): 45.5 -> 34.5Improved
apply Ours-Entropy M=4 without retrainingOurs-Entropy · PA-HMDB51privacy budget accuracy A_B (%): 77.3 -> 71.6Improved
blur or blacken the body instead of the faceempirical obfuscation · SBUprivacy budget accuracy A_B (%): 98.9 -> 21.4Mixed
apply learned anonymization, human ratersf_A · PA-HMDB51 human studymacro-F1 of privacy attribute prediction: 0.97 -> 0.38Improved

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