Easy training on custom dataset. Various backends (MobileNet and SqueezeNet) supported. A YOLO demo to detect raccoon run entirely in brower is accessible at https://git.io/vF7vI (not on Windows).
In order to work as intended, several changes must be made to the coco2pascal.py script that is linked to
first all instances of 'path' must be changed to Path, i.e.
from path import Path (instead of 'from path import path'), then all subsequent calls to Path must be Path() instead of path(). But be careful with wholesale replaces because some other variables use the phrase 'path' in them.
for python3 add parentheses around print statements
make sure all dependencies in header of that file are installed with pip
while in the same folder as the unzipped 'annotations' subfolder, run 4 command like this:
mkdir annote_voc/train
mkdir annote_voc/val
python coco2pascal.py create_annotations ./ val annote_voc/val
python coco2pascal.py create_annotations ./ train annote_voc/train
then set your folders in jupyter notebook accordingly
In order to work as intended, several changes must be made to the coco2pascal.py script that is linked to
first all instances of 'path' must be changed to Path, i.e. from path import Path (instead of 'from path import path'), then all subsequent calls to Path must be Path() instead of path(). But be careful with wholesale replaces because some other variables use the phrase 'path' in them. for python3 add parentheses around print statements make sure all dependencies in header of that file are installed with pip while in the same folder as the unzipped 'annotations' subfolder, run 4 command like this: mkdir annote_voc/train mkdir annote_voc/val python coco2pascal.py create_annotations ./ val annote_voc/val python coco2pascal.py create_annotations ./ train annote_voc/train then set your folders in jupyter notebook accordingly