| import json |
| from pathlib import Path |
|
|
| infile = Path("/data16T/chestx-ray/AURAD_dataset/chexlocalize/test_prompt_text2layout_multi.json") |
| outfile = Path("/data16T/chestx-ray/AURAD_dataset/chexlocalize/test_prompt_text2layout_multi_no_device.json") |
|
|
| removed_attn_count = 0 |
| modified_prompt_count = 0 |
| removed_empty_prompt_count = 0 |
| kept_count = 0 |
|
|
| def remove_support_from_prompt(prompt: str) -> str: |
| if not prompt: |
| return "" |
| parts = [x.strip() for x in prompt.split(",")] |
| new_parts = [x for x in parts if x and "Support Devices" not in x] |
| return ", ".join(new_parts) |
|
|
| with infile.open("r", encoding="utf-8") as f_in, outfile.open("w", encoding="utf-8") as f_out: |
| for line in f_in: |
| line = line.strip() |
| if not line: |
| continue |
|
|
| item = json.loads(line) |
|
|
| old_attn_list = item.get("attn_list", []) |
| new_attn_list = [] |
| for disease, mask_path in old_attn_list: |
| if disease == "Support Devices": |
| removed_attn_count += 1 |
| else: |
| new_attn_list.append([disease, mask_path]) |
| item["attn_list"] = new_attn_list |
|
|
| old_prompt = item.get("prompt", "") |
| new_prompt = remove_support_from_prompt(old_prompt) |
| if new_prompt != old_prompt: |
| modified_prompt_count += 1 |
| item["prompt"] = new_prompt |
|
|
| |
| if item["prompt"] == "": |
| removed_empty_prompt_count += 1 |
| continue |
|
|
| f_out.write(json.dumps(item, ensure_ascii=False) + "\n") |
| kept_count += 1 |
|
|
| print(f"saved to: {outfile}") |
| print(f"kept samples: {kept_count}") |
| print(f"removed 'Support Devices' in attn_list: {removed_attn_count}") |
| print(f"modified prompts: {modified_prompt_count}") |
| print(f"removed empty-prompt samples: {removed_empty_prompt_count}") |
