AI Paper Atlas

Research map / Video / Action recognition

A Unified Framework for Shot Type Classification Based on Subject Centric Lens

Anyi Rao, Jiaze Wang, Linning Xu, Xuekun Jiang, Qingqiu Huang, Bolei Zhou, Dahua Lin

ECCV 2020 · Video › Action recognition · approach: Contrastive video representation learning · 77 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
add subject map guidance to TSN (img) (ablation)TSN-ResNet50 (img) · MovieShots scale typeaccuracy: 84.08 -> 87.21Improved
add subject map guidance to TSN (img) (ablation)TSN-ResNet50 (img) · MovieShots movement typeaccuracy: 70.46 -> 71.30Improved
add variance map to SGNet (img) (ablation)SGNet (img) · MovieShots movement typeaccuracy: 71.30 -> 80.57Improved
add optical flow to SGNet (img) (ablation)SGNet (img) · MovieShots movement typeaccuracy: 71.30 -> 80.65Improved
add variance map and flow to SGNet (img) (ablation)SGNet (img) · MovieShots scale typeaccuracy: 87.21 -> 87.57No clear effect
add Kinetics pre-training to SGNet w/ Var (ablation)SGNet w/ Var (img + flow) · MovieShots movement typeaccuracy: 81.86 -> 83.72Improved
use whole image plus subject branch instead of subject only (ablation)TSN w/ Var img+flow · MovieShots scale typeaccuracy: 79.97 -> 87.57Improved
add background guidance instead of subject guidance for movement (ablation)SGNet w/ Var · MovieShots movement typeaccuracy: 80.86 -> 81.86Improved
add both subject and background guidance (four branches) (ablation)SGNet w/ Var · MovieShots scale typeaccuracy: 87.57 -> 87.31Hurt
replace SBS self-attention with fine-tuned R3Net-ResNet50 subject map (ablation)TSN w/ Var img+flow · MovieShots scale typeaccuracy: 83.82 -> 88.10Improved
add adversarial loss to student generator (ablation)student generator · MovieShots scale typeaccuracy: 85.34 -> 87.08Improved
add adversarial loss to student generator (ablation)student generator · MovieShots movement typeaccuracy: 79.11 -> 81.13Improved

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