''' Input: downloaded datasets Process: resize, change jpg to npy, store images and labels to Image/, Label/ From https://github.com/jcwang123/BA-Transformer ''' import cv2 import os import numpy as np from tqdm import tqdm import matplotlib.pyplot as plt def process_isic2018( isic2018_origin_folder = '/bigdata/siyiplace/data/skin_lesion/2018_raw_data', dim=(512, 512), isic2018_proceeded_folder='/bigdata/siyiplace/data/skin_lesion/isic2018'): # '/raid/wjc/data/skin_lesion/isic2018/') image_dir_path = isic2018_origin_folder+'/ISIC2018_Task1-2_Training_Input/' mask_dir_path = isic2018_origin_folder+'/ISIC2018_Task1_Training_GroundTruth/' # '/raid/wl/2018_raw_data/ISIC2018_Task1_Training_GroundTruth/' image_path_list = os.listdir(image_dir_path) mask_path_list = os.listdir(mask_dir_path) image_path_list = list(filter(lambda x: x[-3:] == 'jpg', image_path_list)) mask_path_list = list(filter(lambda x: x[-3:] == 'png', mask_path_list)) # align masks and inputs image_path_list.sort() mask_path_list.sort() print('number of images: {}, number of masks: {}'.format(len(image_path_list), len(mask_path_list))) # ISBI Dataset for image_path, mask_path in zip(image_path_list, mask_path_list): if image_path[-3:] == 'jpg': print(image_path) assert os.path.basename(image_path)[:-4].split( '_')[1] == os.path.basename(mask_path)[:-4].split('_')[1] _id = os.path.basename(image_path)[:-4].split('_')[1] image_path = os.path.join(image_dir_path, image_path) mask_path = os.path.join(mask_dir_path, mask_path) image = plt.imread(image_path) mask = plt.imread(mask_path) image_new = cv2.resize(image, dim, interpolation=cv2.INTER_CUBIC) mask_new = cv2.resize(mask, dim, interpolation=cv2.INTER_NEAREST) save_dir_path = isic2018_proceeded_folder + '/Image' os.makedirs(save_dir_path, exist_ok=True) np.save(os.path.join(save_dir_path, _id + '.npy'), image_new) save_dir_path = isic2018_proceeded_folder + '/Label' os.makedirs(save_dir_path, exist_ok=True) np.save(os.path.join(save_dir_path, _id + '.npy'), mask_new) def process_PH2( PH2_origin_folder = '/bigdata/siyiplace/data/skin_lesion/PH2_rawdata', PH2_proceeded_folder = '/bigdata/siyiplace/data/skin_lesion/PH2'): PH2_images_path = os.path.join(PH2_origin_folder,'/PH2Dataset/PH2_Dataset_images') path_list = os.listdir(PH2_images_path) path_list.sort() for path in path_list: image_path = os.path.join(PH2_images_path, path, path + '_Dermoscopic_Image', path + '.bmp') label_path = os.path.join(PH2_images_path, path, path + '_lesion', path + '_lesion.bmp') image = plt.imread(image_path) label = plt.imread(label_path) label = label[:, :, 0] dim = (512, 512) image_new = cv2.resize(image, dim, interpolation=cv2.INTER_AREA) label_new = cv2.resize(label, dim, interpolation=cv2.INTER_AREA) image_save_path = os.path.join( PH2_proceeded_folder,'/Image', path + '.npy') # '/data2/cf_data/skinlesion_segment/PH2_rawdata/PH2/Image' label_save_path = os.path.join( PH2_proceeded_folder,'/Label', path + '.npy') # /data2/cf_data/skinlesion_segment/PH2_rawdata/PH2/Label np.save(image_save_path, image_new) np.save(label_save_path, label_new) def process_SKD( SKD_images_folder = '/bigdata/siyiplace/data/skin_lesion/skin_cancer_detection', SKD_proceeded_folder = '/bigdata/siyiplace/data/skin_lesion/SKD'): ''' SKin Cancer Detection dataset ''' SKD_images_path1 = '{}/skin_image_data_set-1/Skin Image Data Set-1/skin_data/melanoma/'.format(SKD_images_folder) SKD_images_path2 = '{}/skin_image_data_set-2/Skin Image Data Set-2/skin_data/notmelanoma/'.format(SKD_images_folder) for images_path in [SKD_images_path1, SKD_images_path2]: for dataset_name in ['dermis', 'dermquest']: path_list = os.listdir('{}{}'.format(images_path, dataset_name)) for path in path_list: if path[-4:] == '.jpg': image_path = os.path.join('{}{}'.format(images_path, dataset_name), path) label_path = os.path.join('{}{}'.format(images_path, dataset_name), path[:-8]+'contour.png') else: continue image = plt.imread(image_path) label = plt.imread(label_path) dim = (512, 512) image_new = cv2.resize(image, dim, interpolation=cv2.INTER_AREA) label_new = cv2.resize(label, dim, interpolation=cv2.INTER_AREA) image_save_path = os.path.join( SKD_proceeded_folder,'/Image', dataset_name+'_'+path[:-4] + '.npy') label_save_path = os.path.join( SKD_proceeded_folder,'/Label', dataset_name+'_'+path[:-4] + '.npy') np.save(image_save_path, image_new) np.save(label_save_path, label_new) def process_DMF( DMF_images_folder = '/bigdata/siyiplace/data/skin_lesion/DMF_origin', DMF_proceeded_folder = '/bigdata/siyiplace/data/skin_lesion/DMF'): ''' Dermofit (DMF) dataset ''' DMF_images_path = '{}/images'.format(DMF_images_folder) path_list = os.listdir(DMF_images_path) path_list.sort() for path in tqdm(path_list): image_path = os.path.join(DMF_images_path, path, path + '.png') label_path = os.path.join(DMF_images_path, path, path + 'mask.png') image = plt.imread(image_path) label = plt.imread(label_path) dim = (512, 512) image_new = cv2.resize(image, dim, interpolation=cv2.INTER_AREA) image_new = np.clip(image_new*255, 0, 255).astype(np.uint8) if image_new.max() < 1.2 else image_new label_new = cv2.resize(label, dim, interpolation=cv2.INTER_AREA) image_save_path = os.path.join( DMF_proceeded_folder,'/Image', path + '.npy') label_save_path = os.path.join( DMF_proceeded_folder,'/Label', path + '.npy') np.save(image_save_path, image_new) np.save(label_save_path, label_new) if __name__ == '__main__': process_isic2018() process_PH2() process_SKD() process_DMF()