Research map / Video / Video segmentation and tracking
Condensing a Sequence to One Informative Frame for Video Recognition
Zhaofan Qiu, Ting Yao, Yan Shu, Chong-Wah Ngo, Tao Mei
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add adversarial regularizer to reconstruction-only IFS (end-to-end frames) (ablation) | IFS app · Kinetics-400 val | top-1 accuracy: 71.7 -> 71.2 | Hurt |
| add adversarial regularizer to reconstruction-only IFS (JPEG frames) (ablation) | IFS app · Kinetics-400 val | top-1 accuracy: 59.2 -> 70.3 | Improved |
| add color consistency regularizer on top of adversarial (ablation) | IFS app+adv · Kinetics-400 val | top-1 accuracy: 71.2 -> 72.8 | Improved |
| add color consistency regularizer on top of adversarial (JPEG frames) (ablation) | IFS app+adv · Kinetics-400 val | top-1 accuracy: 70.3 -> 72.8 | Improved |
| add motion estimation task to reconstruction (ablation) | IFS app · Kinetics-400 val | top-1 accuracy: 72.8 -> 74.5 | Improved |
| add categorization task to reconstruction (ablation) | IFS app · Kinetics-400 val | top-1 accuracy: 72.8 -> 73.7 | Improved |
| add categorization task to app+mot (full IFS) (ablation) | IFS app+mot · Kinetics-400 val | top-1 accuracy: 74.5 -> 75.0 | Improved |
| add appearance reconstruction to cat+mot (ablation) | IFS cat+mot · Kinetics-400 val | top-1 accuracy: 72.8 -> 75.0 | Improved |
| use reconstruction task instead of categorization task alone (ablation) | IFS · Kinetics-400 val | top-1 accuracy: 68.2 -> 72.8 | Improved |
| add appearance reconstruction task to IFS-mot (ablation) | IFS-mot · Kinetics-400 val | top-1 accuracy: 72.8 -> 75.0 | Improved |
| replace dynamic image pooling with learned IFS frame | DI · Kinetics-400 val | top-1 accuracy: 71.6 -> 75.0 | Improved |
| fuse IFS frame with optical flow stream (ablation) | IFS · Kinetics-400 val | top-1 accuracy: 75.0 -> 77.3 | Improved |
Closest papers by meaning
- Video Frame Interpolation without Temporal Priors (NeurIPS 2020 · similarity 0.85)
- Super SloMo: High Quality Estimation of Multiple Intermediate Frames for Video Interpolation (CVPR 2018 · similarity 0.85)
- Stochastic Image-to-Video Synthesis Using cINNs (CVPR 2021 · similarity 0.85)
- Context-Aware Synthesis for Video Frame Interpolation (CVPR 2018 · similarity 0.85)
- IFRNet: Intermediate Feature Refine Network for Efficient Frame Interpolation (CVPR 2022 · similarity 0.85)
- Learning Fine-Grained Features for Pixel-Wise Video Correspondences (ICCV 2023 · similarity 0.85)
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.
