Update Tuberculosis_Dataset.py

Tuberculosis_Dataset.py

@@ -1,11 +1,11 @@
-from datasets import GeneratorBasedBuilder, DownloadManager, DatasetInfo, BuilderConfig, SplitGenerator, Version
+from datasets import GeneratorBasedBuilder, DownloadManager, DatasetInfo, Array3D, BuilderConfig, SplitGenerator, Version
 from datasets.features import Features, Value, Sequence
 import datasets
 import pandas as pd
 import json
 import zipfile
 from PIL import Image
-import os
+import numpy as np
 import io
 
 _DESCRIPTION = """\
@@ -31,7 +31,7 @@ class TuberculosisDataset(GeneratorBasedBuilder):
                 "age": Value("int8"),
                 "case_text": Value("string"),
                 "keywords": Value("string"),
-                "image_filenames": Sequence(Value(dtype="string")),  # Change here to store image filenames
+                "image_arrays": Sequence(Value(dtype="uint8")),
                 "caption": Value("string"),
             }),
             supervised_keys = None,
@@ -47,7 +47,7 @@ class TuberculosisDataset(GeneratorBasedBuilder):
             "caption_json": f"{base_url}image_metadata.json",
             "images_zip": "https://github.com/zhankai-ye/tuberculosis_dataset/raw/main/images/PMC.zip"
         }
-        downloaded_files = dl_manager.download_and_extract(urls)
+        downloaded_files = dl_manager.download(urls)
 
         return [
             SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs=downloaded_files),
@@ -76,10 +76,11 @@ class TuberculosisDataset(GeneratorBasedBuilder):
         merged_df = pd.merge(merged_df, caption[['case_id', 'caption']], on='case_id', how='left')
         merged_df = merged_df.where(pd.notnull(merged_df), None)
         merged_df['age'] = merged_df['age'].astype('int8')
+
         # Prepare images
-        image_filenames = self._prepare_images(images_zip, dl_manager.extracted_path)
+        image_arrays = self._prepare_images(images_zip)
 
-        # Yield examples with minor adjustments for image handling
+        # Yield examples
         for idx, row in merged_df.iterrows():
             yield idx, {
                 "case_id": row["case_id"],
@@ -87,19 +88,20 @@ class TuberculosisDataset(GeneratorBasedBuilder):
                 "age": int(row["age"]),
                 "case_text": row["case_text"],
                 "keywords": row["keywords"],
-                "image_filenames": image_filenames.get(row["case_id"], []),  # Adjusted to yield filenames
+                "image_arrays": image_arrays.get(row["case_id"], []),
                 "caption": row["caption"],
             }
 
-    def _prepare_images(self, images_zip_path, extracted_path):
-        image_filenames = {}
+    def _prepare_images(self, images_zip_path):
+        image_arrays = {}
         with zipfile.ZipFile(images_zip_path, 'r') as zip_ref:
-            zip_ref.extractall(extracted_path)  # Extract images
             for file_info in zip_ref.infolist():
                 if file_info.filename.endswith('.jpg') and not file_info.is_dir():
-                    key = file_info.filename.split('/')[-1].split('_')[0]
-                    if key not in image_filenames:
-                        image_filenames[key] = []
-                    # Save the relative path or identifier of the image instead of its array
-                    image_filenames[key].append(os.path.join(extracted_path, file_info.filename))
-        return image_filenames
+                    with zip_ref.open(file_info) as image_file:
+                        img = Image.open(io.BytesIO(image_file.read()))
+                        img_array = np.array(img)
+                        key = file_info.filename.split('/')[-1].split('_')[0]
+                        if key not in image_arrays:
+                            image_arrays[key] = []
+                        image_arrays[key].append(img_array)
+        return image_arrays