WWTOOL

wwtool is a Python library for Computer Vision Research.

It will provide the following functionalities.

• Basic parse and dump functions for object detection datasets
• Detection annotation visualization
• Dataset convert
• Toolkit for object detection dataset
  • DOTA
  • VisDrone
  • UAV-BD
  • ...

Requirements

• Python 3.7
• Pytorch 1.0+
• CUDA 8.0+
• mmcv

Installation

git clone https://github.com/jwwangchn/wwtool.git
cd wwtool
python setup.py develop

Future works

• Statistic function for object detetion datasets
• Split large image to sub-images
• Results generation tools for dota and visdrone dataset
• Offline data augmentation
• Online data augmentation
• Design structures for bboxes and oriented bboxes

Structure

• scripts
• tests
• tools
• wwtool
  • structures: structures for object detection
  • cnn: CNN Modules by PyTorch
  • models: Models for detection, classification and segmentation
  • ops: C++ or Cython codes for CNN operation
  • datasets: Create and read data from dataset
  • transforms: bbox and image transformation
    • image
    • bbox
  • fileio: file operation
    • image
    • file
    • label
  • visualization: code for visualization
    • image
  • utils: tools for other tasks
    • pid
    • email
    • uart
    • path
    • config
    • progressbar