sec_dir = './wally_dataset/second_dataset'
thr_dir = './wally_dataset/third_dataset'
sec_anns = './wally_dataset/second_dataset/body_crop.csv'
thr_anns = './wally_dataset/third_dataset/body_crop.csv'
#
sec_fgs, sec_bgs = extract_wallybody(sec_dir, sec_anns)
indices = range(len(sec_bgs))
random.shuffle(indices)
thr_fgs, thr_bgs = extract_wallybody(thr_dir, thr_anns)
fgs = np.vstack([sec_fgs, thr_fgs])
bgs = np.vstack([sec_bgs, sec_fgs])
np.save('wally_dataset/wally_imgs.npy', fgs)
np.save('wally_dataset/background_imgs.npy', bgs)
img_prc = ImageProcessing()
train_fg, test_fg, val_fg = img_prc.divide_TVT(fgs, 0.1, 0.1, True)
train_bg, test_bg, val_bg = img_prc.divide_TVT(bgs, 0.1, 0.1, True)
np.save('wally_dataset/wally_train_imgs.npy', train_fg)
np.save('wally_dataset/wally_test_imgs.npy', test_fg)
np.save('wally_dataset/wally_val_imgs.npy', val_fg)
np.save('wally_dataset/background_train_imgs.npy', train_bg)
np.save('wally_dataset/background_test_imgs.npy', test_bg)
np.save('wally_dataset/background_val_imgs.npy', val_bg)
