AI Paper Atlas

Research map / Video / Action recognition

Generic Event Boundary Detection: A Benchmark for Event Segmentation

Mike Zheng Shou, Stan Weixian Lei, Weiyao Wang, Deepti Ghadiyaram, Matt Feiszli

ICCV 2021 · Video › Action recognition · approach: Fully supervised temporal action detection · 75 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
use predictability-based boundary detection instead of random swapping of detectionsPA · TAPOS valF1 avg: 0.314 -> 0.543Improved
use predictability-based unsupervised PA instead of supervised TCNPA · TAPOS valF1 avg: 0.330 -> 0.543Improved
replace PA with supervised PC fine-tuned end-to-endPC · Kinetics-GEBD valF1 avg: 0.527 -> 0.817Improved
replace PA with supervised TCN on fixed featuresTCN · Kinetics-GEBD valF1 avg: 0.527 -> 0.685Improved
train TCN on Kinetics-GEBD instead of TAPOSTCN · Kinetics-GEBD valF1 avg: 0.627 -> 0.685Improved
use BMN-StartEnd probability scores instead of BMN action proposalsBMN · Kinetics-GEBD valF1 avg: 0.223 -> 0.640Improved
use PA instead of shot-change detection SceneDetectPA · Kinetics-GEBD valF1 avg: 0.318 -> 0.527Improved
fine-tune backbone end-to-end in PC vs fixed backbone (ablation)PC · GEBDtraining convergenceImproved
sample middle frame of each segment instead of boundary framesTSN video classification · Kinetics-GEBD valaccuracy (%): 63.0 -> 64.4Improved
sample middle frame of GEBD segments instead of uniform samplingkey frame selection · Kinetics-GEBD val, random samplesvote-level win rate (%): 33.9 -> 40.9Improved
sample middle frame of GEBD segments instead of uniform samplingkey frame selection · Kinetics-GEBD val, large disparityvote-level win rate (%): 12.6 -> 73.0Improved
replace SceneDetect with PA on TAPOSPA · TAPOS valF1 avg: 0.051 -> 0.543Improved

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