Update Populus_Stomatal_Images_Datasets.py

Populus_Stomatal_Images_Datasets.py

@@ -109,30 +109,37 @@ class NewDataset(datasets.GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager):
-        # Only download data, no need to split
+        # Download the CSV file and both ZIP files
         data_files = dl_manager.download_and_extract({
             "csv": "https://huggingface.co/datasets/XintongHe/Populus_Stomatal_Images_Datasets/resolve/main/data/Labeled Stomatal Images.csv",
-            "zip": "https://huggingface.co/datasets/XintongHe/Populus_Stomatal_Images_Datasets/resolve/main/data/Labeled Stomatal Images.zip"
+            "zip_all": "https://huggingface.co/datasets/XintongHe/Populus_Stomatal_Images_Datasets/resolve/main/data/Labeled Stomatal Images.zip",
+            "zip_config": "https://huggingface.co/datasets/XintongHe/Populus_Stomatal_Images_Datasets/resolve/main/data/Labeled Stomatal Images_config.zip"
         })
-        extracted_images_path = 'https://huggingface.co/datasets/XintongHe/Populus_Stomatal_Images_Datasets/resolve/main/data/Labeled Stomatal Images_config.zip'
-
-        # Load the CSV file containing species and scientific names
-        species_info = pd.read_csv(os.path.join(extracted_images_path, 'Labeled Stomatal Images.csv'))
-
-        # Get all image filenames from the new configuration directory
-        all_image_filenames = [f for f in os.listdir(extracted_images_path) if f.endswith('.jpg')]
-
         
+        # Load the CSV file containing species and scientific names
+        species_info_full = pd.read_csv(data_files["csv"])
         
-        # No longer need to randomize and split the dataset
-        return [datasets.SplitGenerator(
-            name=datasets.Split.TRAIN,
-            gen_kwargs={
-                "filepaths": all_image_filenames,
-                "species_info": species_info,
-                "data_dir": extracted_images_path
-            },
-        )]
+        # Assume the config ZIP is extracted to 'Labeled Stomatal Images_config' folder
+        extracted_config_path = os.path.join(dl_manager._cache_dir, "Labeled Stomatal Images_config")
+    
+        # List all image filenames in the extracted config directory
+        all_image_filenames_config = [f for f in os.listdir(extracted_config_path) if f.endswith('.jpg')]
+    
+        # Filter the species_info dataframe to only include the rows with filenames that are in the new configuration
+        # Here, we're extracting the filename without the extension to match with the 'FileName' column in the CSV
+        species_info = species_info_full[species_info_full['FileName'].isin([os.path.splitext(f)[0] for f in all_image_filenames_config])]
+    
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                gen_kwargs={
+                    "filepaths": all_image_filenames_config,
+                    "species_info": species_info,
+                    "data_dir": extracted_config_path
+                },
+            )
+        ]
+
 
 
     def _parse_yolo_labels(self, label_path, width, height):