microsoft / VoTT

Visual Object Tagging Tool: An electron app for building end to end Object Detection Models from Images and Videos.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How to calculate video frame number from VoTT exported image timestamp

haimat opened this issue · comments

When I export tagged frames from a video via VoTT, I get a list of all exported images in a CSV file. Each of these images contain their timestamp from within the video in their file names. For example:

video.mp4#t=260.433333.jpg

Now I want to calculate which frame this image is in the video. So in this example when this video.mp4 has 60 FPS, I would assume that I can simply calculate that frame number via

timestmap * fps = int(260.433333 * 60) = 15626

Thus I would assume that timestamp 260.433333 corresponds to frame number 15626 in the video. As it turned out, however, this is close, but not exactly correct. After some manual tweaking I found out the correct frame number is 15610. The further I go in video time, the greater this difference becomes.

Now I am wondering: Where does this difference of 16 frames come from, and how do I correctly calculate the frame number from these timestamps?

Hello there, any ideas on this issue?