ChronoMagic-Bench Leaderboard
Welcome to the leaderboard of the ChronoMagic-Bench! (NeurIPS 2024 D&B Spotlight)
šChronoMagic-Bench represents the inaugural benchmark dedicated to assessing T2V models' capabilities in generating time-lapse videos that demonstrate significant metamorphic amplitude and temporal coherence. The benchmark probes T2V models for their physics, biology, and chemistry capabilities, in a free-form text control.
If you like our project, please give us a star ā on GitHub for the latest update.
GitHub | arXiv | Home Page | ChronoMagic-Pro | ChronoMagic-ProH
In the table below, we summarize each task performance of all the models. We use UMT-FVD, UMTScore, MTScore, CHScore, GPT4o-MTScore as the primary evaluation metric for each tasks.
U-Net | 257.56 | 1.916 | 0.478 | 81.82 | 3.13 |
In the table below, we summarize each task performance of all the models. We use UMT-FVD, UMTScore, MTScore, CHScore, GPT4o-MTScore as the primary evaluation metric for each tasks.
U-Net | 294.72 | 1.763 | 0.479 | 125.25 | 3.05 |
Submission Guidelines
- Fill in 'Model Name' if it is your first time to submit your result or Fill in 'Revision Model Name' if you want to update your result.
- Select āBackbone Typeā (DiT or U-Net).
- Fill in your home page to 'Model Link'.
- After evaluation, follow the guidance in the github repository to obtain
ChronoMagic-Bench-Input.jsonand upload it here. - Click the 'Submit Eval' button.
- Click 'Refresh' to obtain the uploaded leaderboard.