Everything outside the wells is now set to 0 in the masked videos. Namely, I find the left edge of the leftmost wells, the right edge of the rightmost well, the top edge of the topmost well, the bottom edge of the bottommost well. Every pixel respectively to the right, left, top, and bottom of these limits, is set to 0.
Brings down file size.
NN filtering of blobs
Including Ziwei's latest pytorch model, this is the new pytorch_default.
Bugfix
In get_tierpsy_features.py, the function save_feats_stats was incorrectly detecting MWP videos with no worms but dirt outside the wells as non-MWP videos, so producing features_stats for said dirt. Now the test of whether a video is of a MWP is a check that /fov_wells exists in the *_featuresN.hdf5 file.
FOV splitter
Everything outside the wells is now set to 0 in the masked videos. Namely, I find the left edge of the leftmost wells, the right edge of the rightmost well, the top edge of the topmost well, the bottom edge of the bottommost well. Every pixel respectively to the right, left, top, and bottom of these limits, is set to 0. Brings down file size.
NN filtering of blobs
Including Ziwei's latest pytorch model, this is the new
pytorch_default.Bugfix
In
get_tierpsy_features.py, the functionsave_feats_statswas incorrectly detecting MWP videos with no worms but dirt outside the wells as non-MWP videos, so producingfeatures_statsfor said dirt. Now the test of whether a video is of a MWP is a check that/fov_wellsexists in the*_featuresN.hdf5file.