Nightmare-n / DepthAnyVideo

Depth Any Video with Scalable Synthetic Data
https://depthanyvideo.github.io
Apache License 2.0
363 stars 24 forks source link

Inference on longer videos #14

Open evgeniishch opened 1 week ago

evgeniishch commented 1 week ago

Dear authors, thank you for your work!

What is the suggested usage for videos of arbitrary length? Say, >1000 frames?

Thank you!

Nightmare-n commented 1 week ago

We have not experimented with videos exceeding 1000 frames. However, one possible approach could be to sample key frames from the sequence and then recursively apply a frame interpolation network to reconstruct the intermediate frames.