Research map / Video / Video segmentation and tracking
MBTI: Masked Blending Transformers with Implicit Positional Encoding for Frame-rate Agnostic Motion Estimation
Jungwoo Huh, Yeseung Park, Seongjean Kim, Jungsu Kim, Sanghoon Lee
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 masked token blending to the MAE baseline (ablation) | MAE baseline · EMDB-FPS Split 2 (global) | MCF: 22.5 -> 15.9 | Improved |
| add implicit positional encoding to the MAE baseline (ablation) | MAE baseline · EMDB-FPS Split 2 (global) | MCF: 22.5 -> 19.2 | Improved |
| combine implicit PE with masked token blending (ablation) | MBTI · EMDB-FPS Split 2 (global) | MCF: 15.9 -> 6.7 | Improved |
| replace gamma-only time encoding with neural implicit function (ablation) | MBTI w/ MTB · EMDB-FPS Split 2 (global) | MCF: 9.7 -> 6.7 | Improved |
| add implicit PE and MTB (ablation) | MAE baseline · EMDB-FPS Split 2 (global) | Jitter std: 49.4 -> 41.7 | Improved |
| add implicit PE and MTB (ablation) | MAE baseline · EMDB-FPS Split 2 (global) | W-MPJPE mean: 341.7 -> 224.4 | Improved |
| train TRAM on frame-rate-varied data (ablation) | TRAM · EMDB-FPS Split 1 (local) | PA-MPJPE / MPJPE at 30 FPS: PA-MPJPE up 2.3, MPJPE up 3.8 | Hurt |
| train MBTI on frame-rate-varied data instead of fixed 30 FPS (ablation) | MBTI · EMDB-FPS Split 1 (local) | MCF: 7.8 -> 7.4 | Improved |
| train MBTI on frame-rate-varied data instead of fixed 30 FPS (ablation) | MBTI · EMDB-FPS Split 1 (local) | PA-MPJPE / MPJPE at 30 FPS: PA-MPJPE 2.8 higher, MPJPE 4.8 higher | Hurt |
| use MBTI instead of TRAM | TRAM · EMDB-FPS Split 1 (local) | MCF: 8.6 -> 7.4 | Improved |
| use MBTI instead of GVHMR | GVHMR · EMDB-FPS Split 2 (global) | MCF: 25.6 -> 6.7 | Improved |
| use MBTI instead of TRAM on 3DPW | TRAM · 3DPW | PA-MPJPE: 35.6 -> 35.0 | Improved |
Closest papers by meaning
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.87)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.86)
- MotionPro: A Precise Motion Controller for Image-to-Video Generation (CVPR 2025 · similarity 0.86)
- Seamless Human Motion Composition with Blended Positional Encodings (CVPR 2024 · similarity 0.86)
- MotionAura: Generating High-Quality and Motion Consistent Videos using Discrete Diffusion (ICLR 2025 · similarity 0.86)
- MotionStone: Decoupled Motion Intensity Modulation with Diffusion Transformer for Image-to-Video Generation (CVPR 2025 · 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.
