search

yeephycho / tensorflow-face-detection

A mobilenet SSD based face detector, powered by tensorflow object detection api, trained by WIDERFACE dataset.
Apache License 2.0
768 stars 264 forks source link

Why frame_num = 1490 ? #20

Closed shalabhsingh closed 6 years ago

shalabhsingh commented 6 years ago

Hi, I was going through your source code but couldn't understand why you have used frame_num = 1490 and run through only these many frames in inference_video_face.py file? Can you please clarify it?

yeephycho commented 6 years ago

Hi @shalabhsingh, 1490 is my lucky number. You can change with your will.

shalabhsingh commented 6 years ago

Hi, my question is why are you limiting the frame numbers? If a video has more than 1490 frames, will it get processed correctly? If yes how? If no, why have you restricted the frame numbers?

xiangyann commented 6 years ago

The excess frames will not be processed, 1490 means that only the first minute will be parsed 1490f/25fps = 59.6 sec Just assign a large number for whole/long video processing 1 minute is useful for demonstration purposes IMO.