I decided to participate in the Kaggle Competition. The target of this time is only fish, and the idea is that it may be possible to use a method other than Deep. But is it still Faster RCNN that wins ...
conda install opencv
-Face detection --Not training -Cat detection --I am also training. It's wonderful that you also collect images of cats yourself.
https://github.com/mrnugget/opencv-haar-classifier-training
Thank you for using. Basically, follow the instructions in the README.
Ignore "1." because the environment has already been constructed. Start with "2."
For "3." and "4.", be sure to add the -follow option if the positive_images folder is a symbolic link.
Put the cropped fish image in positive_images (Kaggle's dataset is not cropped, so you need to do your best to crop it), and put the image without fish in negative_images and execute the following command!
find ./positive_images -follow -iname "*.jpg " > positives.txt
If an error occurs, "6." can be solved with the following command.
find . -empty | xargs rm -rf
Recommended Posts