yoloproc 0.1.0

No project description provided

yoloproc

A Python package for processing YOLO datasets.

Features

yoloproc provides a collection of tools to manage and preprocess YOLO-formatted datasets. The main features include:

1. Extracting Datasets from Archives: Supports extraction of .zip and .7z archives.

2. Merging YOLO Data: Merges images and labels from multiple subdirectories into unified images and labels folders.

3. Removing Irrelevant Data: Cleans datasets by removing images without labels and vice versa.

4. Extracting Ordered Frames: Extracts frames from image sequences at a specified interval.

5. Generating Empty Annotations: Creates empty label files for images without corresponding annotations.

6. Analyzing Label Information: Counts the number of instances for each class in label files.

7. Organizing Labels: Organizes labels and images into class-specific directories.

8. Visualizing Labeled Images: Draws bounding boxes on images according to YOLO annotations.

9. Rewriting Labels: Rewrites label files based on a new class mapping.

10. Cropping Dataset Images and Labels: Crops images and adjusts labels accordingly.

11. Converting YOLO to COCO Format: Converts a YOLO-formatted dataset to COCO format.

Installation

git clone https://github.com/yourusername/yoloproc.git cd yoloproc pip install .

Usage

1. Extracting Datasets from Archives

Extract all supported archives in a directory to a specified output directory.

from yoloproc.archive_extractor import ArchiveExtractor

extractor = ArchiveExtractor()
extractor.extract_all(
    source_dir='/path/to/archives',
    output_dir='/path/to/output'
)

2. Merging YOLO Data

Merge images and labels scattered across subdirectories into unified images and labels folders.

from yoloproc.script import merge_yolo_data

merge_yolo_data(
    root_path='/path/to/dataset',
    output_path='/path/to/output',
    is_move=False  # Set to True to move files instead of copying
)

3. Removing Irrelevant Data

Remove files without corresponding pairs (e.g., images without labels).

from yoloproc.script import remove_irrelevant_data

# Remove label files (*.txt) without corresponding images
remove_irrelevant_data(
    root_path='/path/to/dataset',
    remove_type='*.txt'
)

# Remove image files (*.jpg) without corresponding labels
remove_irrelevant_data(
    root_path='/path/to/dataset',
    remove_type='*.jpg'
)

4. Extracting Ordered Frames

Extract frames from a sequence of images at a specified interval.

from yoloproc.script import extract_ordered_frames

extract_ordered_frames(
    image_dir='/path/to/images',
    output_dir='/path/to/output',
    frame_interval=10,  # Extract every 10th frame
    label_dir='/path/to/labels'  # Optional: handle corresponding labels
)

5. Generating Empty Annotations

Create empty label files for images without annotations.

from yoloproc.script import generate_empty_annotations

generate_empty_annotations(
    image_dir='/path/to/images',
    label_dir='/path/to/labels'
)

6. Analyzing Label Information

Count instances of each class in label files.

from yoloproc.script import labels_info

labels_info(
    labels_path='/path/to/labels',
    class_path='/path/to/classes.txt'  # Optional
)

7. Organizing Labels

Organize labels and images into class-specific directories.

from yoloproc.yolo_visual_cls import LabelOrganizer

organizer = LabelOrganizer(
    dataset_root='/path/to/dataset',
    class_file='/path/to/classes.txt'  # Optional
)
organizer.organize_labels(
    is_move=False  # Set to True to move files instead of copying
)

8. Visualizing Labeled Images

Draw bounding boxes on images based on YOLO annotations.

from yoloproc.yolo_im_visual import create_visualizer

create_visualizer(
    labels_path='/path/to/labels',
    images_path='/path/to/images',       # Optional
    visuals_path='/path/to/visuals',     # Optional
    class_path='/path/to/classes.txt',   # Optional
    interval=1,                          # Process every image
    backend='yolo'                       # 'yolo' or 'mpl' for different visualization styles
)

9. Rewriting Labels

Rewrite label files with a new class mapping.

from yoloproc.label_rewrite import rewrite_yolo_labels

# Define a new class mapping
class_filter = {
    0: 0,  # Old class 0 mapped to new class 0
    1: 1,  # Old class 1 mapped to new class 1
    2: 2   # Old class 2 mapped to new class 2
}

rewrite_yolo_labels(
    labels_path='/path/to/labels',
    class_filter=class_filter,
    output_dir='labels2',
    remove_empty=False,
    custom_output_path='/path/to/output/labels'  # Optional
)

10. Cropping Dataset Images and Labels

Crop images and adjust corresponding labels.

from yoloproc.dataset_crop import DatasetCropGestures

crop_origin = (480, 284, 1440, 796)  # Define crop coordinates
dataset_path = '/path/to/dataset'
output_path = '/path/to/output'

dataset_crop = DatasetCropGestures(
    dataset_path=dataset_path,
    output_path=output_path,
    crop_origin_xyxy=crop_origin
)
dataset_crop.crop_all_thread()

11. Converting YOLO to COCO Format

Convert a YOLO-formatted dataset to the COCO annotation format.

from yoloproc.yolo2coco import convert_yolo_to_coco

convert_yolo_to_coco(
    image_dir='/path/to/dataset',
    yolo_class_txt='/path/to/classes.txt'  # Optional
)

Dependencies

• Python 3.9 or higher
• tqdm for progress bars
• opencv-python for image processing
• numpy for numerical operations
• Pillow for image handling
• matplotlib for visualization (optional)
• rich for enhanced terminal output (optional)

License

This project is licensed under the MIT License. See the LICENSE file for details.

Author

• wm - Initial work - Email

Acknowledgments

• Thanks to all contributors and open-source projects that have provided inspiration and code examples.

Contributing

Contributions are welcome! Please open an issue or submit a pull request on GitHub.

TODO

• Add support for additional archive formats.
• Improve error handling and logging.
• Expand test coverage and add more examples.
• Add dataset splitting functionality (train/val/test)