| import os
|
| import numpy as np
|
| from PIL import Image
|
| from datasets import Dataset, DatasetDict, GeneratorBasedBuilder, SplitGenerator, DatasetInfo, Image, Features, Image, Value, Sequence, Split
|
|
|
|
|
| class YoloDataset(GeneratorBasedBuilder):
|
|
|
| def _info(self):
|
| return DatasetInfo(
|
| description="YOLO-style dataset",
|
| features=Features({
|
| 'image': Image(),
|
| 'label': Sequence({
|
| 'class_id': Value('int32'),
|
| 'x_center': Value('float32'),
|
| 'y_center': Value('float32'),
|
| 'width': Value('float32'),
|
| 'height': Value('float32'),
|
| }),
|
| }),
|
| supervised_keys=None,
|
| homepage="https://huggingface.co/datasets/your_username/your_dataset_name",
|
| license="MIT",
|
| )
|
|
|
| def _split_generators(self, dl_manager):
|
| data_dir = os.path.join(self.config.data_dir, "data")
|
| return [
|
| SplitGenerator(
|
| name=Split.TRAIN,
|
| gen_kwargs={"images_path": os.path.join(data_dir, "images/train"),
|
| "labels_path": os.path.join(data_dir, "labels/train")}
|
| ),
|
| SplitGenerator(
|
| name=Split.VALIDATION,
|
| gen_kwargs={"images_path": os.path.join(data_dir, "images/val"),
|
| "labels_path": os.path.join(data_dir, "labels/val")}
|
| ),
|
| ]
|
|
|
| def _generate_examples(self, images_path, labels_path):
|
| """Yields examples."""
|
| for image_file in os.listdir(images_path):
|
| if not image_file.endswith((".jpg", ".png", ".jpeg")):
|
| continue
|
|
|
| image_path = os.path.join(images_path, image_file)
|
| label_file = os.path.splitext(image_file)[0] + ".txt"
|
| label_path = os.path.join(labels_path, label_file)
|
|
|
| objects = []
|
| if os.path.exists(label_path):
|
| with open(label_path, "r", encoding="utf-8") as f:
|
| for line in f:
|
| parts = line.strip().split()
|
| if len(parts) == 5:
|
| class_id, x_center, y_center, width, height = map(float, parts)
|
| objects.append({
|
| "class_id": int(class_id),
|
| "x_center": x_center,
|
| "y_center": y_center,
|
| "width": width,
|
| "height": height
|
| })
|
|
|
| yield image_file, {"image": image_path, "label": objects}
|
|
|