from glob import glob import os import json import random train_cnt = 0 test_cnt = 0 val_cnt = 0 combined_cnt = 0 train_final = {} test_final = {} val_final = {} combined_final = {} for fn in glob("data/tasks/*/*"): print(fn) app_name = fn.split('/')[3] app_type = fn.split('/')[2] for i in range(1,10): image = f"data/data/{app_type}/{app_name}/screen_{i}.png" if os.path.exists(image) == False: if i == 1: print(image) continue ocr = f"ocr/{app_type}/{app_name}/screen_{i}.json" assert os.path.exists(ocr), ocr color = f"detact_color/{app_type}/{app_name}/screen_{i}.json" assert os.path.exists(color), color icon = f"detact_icon/{app_type}/{app_name}/screen_{i}.json" assert os.path.exists(icon), icon box = f"data/metadata/{app_type}/boxes/{app_name}/screen_{i}.json" assert os.path.exists(box), box script_to_task_fn = {} cnt = 0 for task_fn in glob(f"{fn}/task_{i}.*.txt"): txt = '\n'.join(open(task_fn).read().strip().split('\n')[1:]) if txt not in script_to_task_fn: script_to_task_fn[txt] = [] script_to_task_fn[txt].append(task_fn) cnt += 1 for k in script_to_task_fn: c_tasks = script_to_task_fn[k] rand = random.randint(1,10) if rand <= 7: for c in c_tasks: train_final[train_cnt] = {"task": c, "image": image, "ocr": ocr, "color": color, "icon": icon, "box": box} train_cnt += 1 combined_final[combined_cnt] = {"task": c, "image": image, "ocr": ocr, "color": color, "icon": icon, "box": box} combined_cnt += 1 elif rand == 8: for c in c_tasks: val_final[val_cnt] = {"task": c, "image": image, "ocr": ocr, "color": color, "icon": icon, "box": box} val_cnt += 1 combined_final[combined_cnt] = {"task": c, "image": image, "ocr": ocr, "color": color, "icon": icon, "box": box} combined_cnt += 1 elif rand > 8: for c in c_tasks: test_final[test_cnt] = {"task": c, "image": image, "ocr": ocr, "color": color, "icon": icon, "box": box} test_cnt += 1 combined_final[combined_cnt] = {"task": c, "image": image, "ocr": ocr, "color": color, "icon": icon, "box": box} combined_cnt += 1 print(train_cnt) print(val_cnt) print(test_cnt) print(combined_cnt) json.dump(train_final, open('train.json', "w"), indent=4) json.dump(val_final, open("val.json","w"), indent=4) json.dump(test_final, open("test.json","w"), indent=4) json.dump(combined_final, open('combined.json','w'), indent=4)