useful tools for Deep Learning & Computer Vision projects
draw_graphs - If you import this file when training, it will show you accuracy&loss graph dynamically. You can also make yyplot and confusion matrix.
extractor - Extract an arbitary ratio images from assigned image folder randomly.
image_resize_LANCZOS - create new image folder inside new "resize" directory, having resized images.
path_changer - change path inside csv to fit your environment.
train_test_separater - from folder which has imgs for each class, create dataset which has 20% test imgs and 80% train imgs, which is randomly selected from original folder.
integrate_img4dataset - this is designed to use after train_test_separater.py. it will collect dataset which has designated suffix and integrate them to make one big dataset.
video2img - from a video, it will create folder having every frame img.
write_gspread - it will write final DL epoch result on google spreadsheet.
video2img --you prepare some videos for image dataset.
↓
integrate_img4dataset -- With various imgs from some videos, create one big img folder for dataset.
↓
train_test_separater -- If you have created classes and sorted imgs above dataset accordingly, this script will make train&test folder automatically.
↓
image_resize_LANCZOS -- If you wanna change image size, you can use this b4 training.
If you wanna make dataset for object-counting, you can refer to this repository: Count-Annotator2.
draw_graphs -- you can check accuracy&loss in real time.
write_gspread -- you can check current learning status even on your smartphone!