| import torch |
| import os |
| import json |
| import gc |
| import csv |
|
|
| from PIL import Image |
| from datasets import load_dataset |
| from collections import defaultdict |
| from tqdm import tqdm |
|
|
| import numpy as np |
|
|
| from ..utils import save_json,extract |
| from ..base_dataset import BaseDataset |
|
|
| from ..question_formats import get_report_generation_prompt |
|
|
| class CheXpert_Plus(BaseDataset): |
| def __init__(self,model,dataset_path,output_path): |
| self.model = model |
| self.output_path = output_path |
| self.dataset_path = dataset_path |
| self.samples = [] |
| self.chunk_idx = int(os.environ.get("chunk_idx",0)) |
| self.num_chunks = int(os.environ.get("num_chunks",1)) |
| |
| def load_data(self): |
| self.maybe_download_dataset() |
| dataset_path = self.dataset_path |
| json_path = os.path.join(dataset_path,"test.json") |
|
|
| with open(json_path,"r") as f: |
| dataset = json.load(f) |
|
|
| for idx,sample in tqdm(enumerate(dataset)): |
| if idx % self.num_chunks == self.chunk_idx: |
| if sample["findings"].strip() == "" and sample["impression"].strip() == "": |
| continue |
| sample = self.construct_messages(sample) |
| self.samples.append(sample) |
| print("total samples number:", len(self.samples)) |
| return self.samples |
|
|
| def construct_messages(self,sample): |
| image_root = os.path.join(self.dataset_path,"images") |
| image = sample["image"] |
| image = Image.open(os.path.join(image_root,image)) |
| findings = sample["findings"] |
| impression = sample["impression"] |
|
|
| findings = "None" if findings.strip() == "" else findings |
| impression = "None" if impression.strip() == "" else impression |
| |
| prompt = get_report_generation_prompt() |
| messages = {"prompt":prompt,"image":image} |
| sample["messages"] = messages |
| return sample |
|
|
|
|
| def cal_metrics(self,out_samples): |
| import pandas as pd |
|
|
| predictions_data = [] |
| ground_truth_data = [] |
|
|
| for i,sample in enumerate(out_samples): |
| response = sample["response"] |
| findings = sample["findings"] |
| impression = sample["impression"] |
| golden = f"Findings: {findings} Impression: {impression}." |
|
|
| |
| study_id = f"study_{i+1}" |
| |
| |
| predictions_data.append({ |
| 'study_id': study_id, |
| 'report': response |
| }) |
|
|
| |
| ground_truth_data.append({ |
| 'study_id': study_id, |
| 'report': golden |
| }) |
|
|
|
|
| |
| predictions_df = pd.DataFrame(predictions_data) |
| ground_truth_df = pd.DataFrame(ground_truth_data) |
|
|
| prediction_path = os.path.join(self.output_path,'predictions.csv') |
| ground_truth_path = os.path.join(self.output_path,'ground_truth.csv') |
| |
| predictions_df.to_csv(prediction_path, index=False) |
| ground_truth_df.to_csv(ground_truth_path, index=False) |
|
|
| return {"total metrics":"please use cal_report_metrics.py to generate metrics"},out_samples |
| |
|
|
| def maybe_download_dataset(self): |
| if not os.path.exists(self.dataset_path): |
| raise ValueError(f"Dataset path {self.dataset_path} does not exist. Please download from https://aimi.stanford.edu/datasets/chexpert-plus and put it under {self.dataset_path}.") |
| |
