JunweiLiang / Object_Detection_Tracking

Out-of-the-box code and models for CMU's object detection and tracking system for multi-camera surveillance videos. Speed optimized Faster-RCNN model. Tensorflow based. Also supports EfficientDet. WACVW'20
MIT License
486 stars 134 forks source link

Issue running sample command. #9

Closed melkiga closed 4 years ago

melkiga commented 4 years ago

Hi! I'm trying to execute your updated code on a single video file. I modified the paths in your sample command to suit the ones on my machine and I get an Assertion error that looks like the following: Traceback (most recent call last): File "./obj_detect_tracking_cmu_new/obj_detect_tracking.py", line 429, in <module> vcap = VideoReader(videofile) File "/home/gmelki1/activity-detection/argus-images/obj_detect_tracking_cmu_new/diva_io/video/reader.py", line 53, in __init__ self._init() File "/home/gmelki1/activity-detection/argus-images/obj_detect_tracking_cmu_new/diva_io/video/reader.py", line 219, in _init assert len(self._packets) == self._stream.duration AssertionError

Any ideas? Thanks!

JunweiLiang commented 4 years ago

I need more information. This looks like your video is corrupted.

  1. What's the output of ffprobe of your video?
  2. Does opencv frame extraction work? (run without --use_lijun)
  3. Does MoviePy frame extraction work? (run without --use_lijun and with --use_moviepy)

If you are sure that your video is not corrupted, put the video online somewhere so @Lijun-Yu can test it.

Lijun-Yu commented 4 years ago

This is a known issue of diva_io v0.2 on some corrupted videos, which has been fixed in diva_io v0.3. I would recommend @JunweiLiang to update correspondingly to fix this.

JunweiLiang commented 4 years ago

@melkiga Please add the three outputs I mentioned above.

melkiga commented 4 years ago

Thanks for the response. I was able to get it running by running the command without --use_lijun. I'm positive the video is not corrupted.

JunweiLiang commented 4 years ago

FYI, the code is updated with diva_io v0.3 now.

melkiga commented 4 years ago

Thanks, appreciate the heads up.