AI Paper Atlas

Research map / Video / Video-language understanding

Character Grounding and Re-Identification in Story of Videos and Text Descriptions

Youngjae Yu, Jongseok Kim, Heeseung Yun, Jiwan Chung, Gunhee Kim

ECCV 2020 · Video › Video-language understanding · approach: Video-language pretraining · 6 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 body-part embedding to face-only video re-id (ablation)CiSIN · M-VAD Names video re-idaccuracy: 0.783 -> 0.799Improved
add simple average-pooled human bbox embedding to face (ablation)CiSIN · M-VAD Names video re-idaccuracy: 0.783 -> 0.781No clear effect
add text context (grounding and text matching) to video re-id (ablation)CiSIN · M-VAD Names video re-idaccuracy: 0.799 -> 0.806Improved
add face embedding for re-id vs ImageNet ResNet-50CiSIN · M-VAD Names video re-idaccuracy: 0.617 -> 0.783Improved
add motion embedding (ablation)CiSIN · M-VAD Names groundingaccuracy: 0.606 -> 0.684Improved
add attribute (gender) module and Text Re-Id joint training (ablation)CiSIN · M-VAD Names groundingaccuracy: 0.651 -> 0.673Improved
add Text Re-Id Graph joint training (ablation)CiSIN · M-VAD Names groundingaccuracy: 0.673 -> 0.684Improved
add attribute loss and grounding graph to BERT text re-id (ablation)CiSIN · M-VAD Names text re-idaccuracy: 0.734 -> 0.743Improved
add Character Grounding graph training (ablation)CiSIN · M-VAD Names text re-idaccuracy: 0.743 -> 0.754Improved
add Video Re-Id graph (full joint training) (ablation)CiSIN · M-VAD Names text re-idaccuracy: 0.754 -> 0.767Improved
add Text Re-Id graph (ablation)CiSIN · M-VAD Names text re-idaccuracy: 0.737 -> 0.767Improved
use joint training instead of separate training (ablation)CiSIN · LSMDC 2019 Fill-in the Charactersaccuracy: 0.653 -> 0.673Improved

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