Value error with ROI location

[syadavno3]

We consistently encounter an issue when generating the CSV file for our individual tracking video data. We don't exactly know what the issue occurring is and we have made sure that our regions are set correctly and that the video is not corrupt. The entire video was covered by only two defined regions: left and right. Another problem affecting other videos and their CSV files was the omission of the "right" region, showing only "left" and "non-roi". We are not looking for transition data at all and therefore have no empty space on the frames. Over the 10 videos that I have processed, these issues have only occurred in about 3 of them.

[ZachPenn]

This is not an issue generating the CSV, but rather, when defining ROI positions after tracking the animals center of mass for the entire video.

Can you re-run the example you gave - including the following lines of code before defining location in the second line - and share your output?

print(video_dict['region_names'])
print(video_dict['roi_stream'])

[syadavno3]

This is what I got

[ZachPenn]

Thanks. I'm guessing the problem is the result of having the X,Y coordinates of the animal on some frames being assigned to nan. When this is the case, an error is thrown when trying to define which ROIs the animal is in. I will eventually address this. In the interim, I would modify your tracking parameters so that you don't have the positional coordinates be nans. In short, you need to make your tracking parameters less stringent. You can decrease loc_thresh parameter. If you are using the wire removal algorithm, consider reducing the kernel size.

Let me know if this helps.