waste_dumpsites.py

import json
import os
from pathlib import Path
import datasets

class WasteDumpsitesConfig(datasets.BuilderConfig):
    def __init__(self, **kwargs):
        super().__init__(version=datasets.Version("1.0.0"), **kwargs)

class WasteDumpsites(datasets.GeneratorBasedBuilder):
    """Waste Dumpsites Drone Imagery dataset in COCO format."""

    BUILDER_CONFIGS = [
        WasteDumpsitesConfig(name="default", description="Default config for Waste Dumpsites dataset")
    ]

    def _info(self):
        print("datasets.DatasetInfo")
        return datasets.DatasetInfo(
            features=datasets.Features({
                "image": datasets.Image(),
                "image_id": datasets.Value("int64"),
                "width": datasets.Value("int64"),
                "height": datasets.Value("int64"),
                "annotations": datasets.Sequence({
                    "id": datasets.Value("int64"),
                    "bbox": datasets.Sequence(datasets.Value("float32"), length=4),
                    "area": datasets.Value("float32"),
                    "category_id": datasets.Value("int64"),
                    "category_name": datasets.Value("string"),
                    "iscrowd": datasets.Value("int64"),
                })
            }),
            supervised_keys=None,
        )

    def _split_generators(self, dl_manager):
        """Returns SplitGenerators."""
        print("Returns SplitGenerators.")
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"split": "train"}
            ),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION,
                gen_kwargs={"split": "valid"}
            ),
            datasets.SplitGenerator(
                name=datasets.Split.TEST,
                gen_kwargs={"split": "test"}
            ),
        ]

    def _generate_examples(self, split):
        """Yields examples for COCO format."""
        print("Yields examples for COCO format.")
        # Using your exact directory and file structure
        split_dir = os.path.join(self.config.data_dir, split)
        annotations_file = os.path.join(split_dir, "annotations_coco.json")
        
        # Read COCO format annotations
        with open(annotations_file, 'r') as f:
            coco_data = json.load(f)
            
        # Create category id to name mapping
        categories = {cat['id']: cat['name'] for cat in coco_data['categories']}
        
        # Group annotations by image
        image_annotations = {}
        for ann in coco_data['annotations']:
            img_id = ann['image_id']
            if img_id not in image_annotations:
                image_annotations[img_id] = []
            image_annotations[img_id].append(ann)
        
        # Yield examples
        for image in coco_data['images']:
            image_id = image['id']
            print(image_id)
            
            # Get all annotations for this image
            annotations = image_annotations.get(image_id, [])
            
            # Prepare annotations in the required format
            formatted_annotations = {
                "id": [],
                "bbox": [],
                "area": [],
                "category_id": [],
                "category_name": [],
                "iscrowd": []
            }
            
            for ann in annotations:
                formatted_annotations["id"].append(ann["id"])
                formatted_annotations["bbox"].append(ann["bbox"])
                formatted_annotations["area"].append(ann["area"])
                formatted_annotations["category_id"].append(ann["category_id"])
                formatted_annotations["category_name"].append(categories[ann["category_id"]])
                formatted_annotations["iscrowd"].append(ann.get("iscrowd", 0))
            
            yield str(image_id), {
                "image": os.path.join(split_dir, image["file_name"]),
                "image_id": image_id,
                "width": image["width"],
                "height": image["height"],
                "annotations": formatted_annotations
            }