yolodt 0.4.0

YOLO Dataset Tools - Comprehensive toolkit for YOLO format dataset processing

YDT - YOLO Dataset Tools

English | 简体中文

Features

• Auto-detects and handles both OBB (9 values: class_id x1 y1 x2 y2 x3 y3 x4 y4) and BBox (5 values: class_id x_center y_center width height) formats
• SAHI-powered smart slicing for large images with horizontal/grid modes and configurable overlap
• Resize (scale & crop) with custom interpolation (linear/lanczos4),image or yolo dataset
• Coordinate-based precision cropping
• Object cropping from model inference or dataset labels with padding and size filters
• Video frame extraction with parallel processing support
• Smart train/val split with class balancing
• Multi-dataset merging
• Dataset extraction by class IDs with optional label filtering and ID remapping
• Synthetic dataset generation with configurable objects per image, rotation ranges, and balanced class sampling
• YOLO auto-labeling with BBox/OBB format support
• Interactive dataset browser with keyboard controls (n/p/q)

Installation

pip install yolodt

Usage

ydt --help

usage: ydt [-h] [--version] [-v]
           {slice,augment,video,crop-coords,resize,concat,split,merge,extract,synthesize,auto-label,analyze,visualize,viz-letterbox}
           ...

YOLO Dataset Tools - Process and manage YOLO format datasets

positional arguments:
  {slice,augment,video,crop-coords,crop,resize,concat,split,merge,extract,synthesize,auto-label,analyze,visualize,viz-letterbox}
                        Available commands
    slice               Slice large images into tiles
    augment             Augment dataset with rotations
    video               Extract frames from videos
    crop-coords         Crop images by coordinates
    crop                Crop objects from images using model or dataset labels
    resize              Resize images or YOLO dataset
    concat              Concatenate two images
    split               Split dataset into train/val
    merge               Merge multiple datasets
    extract             Extract classes, images, or labels
    synthesize          Generate synthetic dataset
    auto-label          Auto-label images using YOLO model
    analyze             Analyze dataset statistics
    visualize           Visualize YOLO dataset interactively
    viz-letterbox       Visualize letterbox transformation

options:
  -h, --help            show this help message and exit
  --version             show program's version number and exit
  -v, --verbose         Verbose output

🙏 Acknowledgments

• Ultralytics - YOLO framework
• SAHI - Slicing aided hyper inference
• Albumentations - Image augmentation

---