Document

Ego-view Accident Video Diffusion Evaluation

This is our ego-view accident video generation benchmark that can be driven by different text descriptions annotated in MM-AU. The performance is measured by the CLIP Score (CLIPs), Fréchet Video Distance (FVD) and Frames Per Second (FPS). We aim to explore the cause-effect evolution of accident videos conditioned by the descriptions of accident reasons or prevention advice.

ID	Method	Year	Code	CLIPs	FVD	FPS	Runtime	Environment
1	Tune-A-Video	2023	code	21.77	9545.6	1.7	...	GeForce RTX 3090
2	ControlVideo	2023	code	22.51	12275.2	0.5	...	GeForce RTX 3090
3	OVAD	2023.	code	27.24	5238.1	1.2	...	GeForce RTX 3090
4	ModelScope T2V	2023	code	27.15	5088.778	1.3	...	GeForce RTX 3090
5	Text2Video-Zero	2023	code	27.89	12547.075	1.1	...	GeForce RTX 3090

NOTE: You can submit your metric values via the provided form. Furthermore, we would highly appreciate for your contribution with clear links to relevant articles and code for more in-depth analysis.

the result please submit here:

ID	Method	Year	CLIPs	FVD	FPS	Runtime	Environment	Code	Paper