Only detecting image?

[lilun-lll]

Does it detect video?

[joonjeon]

Well, as long as detecting video means basically detecting objects in frames of a video, I think this is at least technically possible.

However, according to official KITTI evaluation web page, it takes more than 3 seconds per frame to be processed... so this model might not be suitable for practical use in frame-by-frame video parsing or detection from real-time scenery.