BSTOOL

bstool is a Python library for Building Segmentation.

It will provide the following functionalities.

• Basic parse and dump functions for building segmentation dataset
• Evaluation tools
• Visualization tools
• Dataset convert

Requirements

• Python 3.6+
• Pytorch 1.1+
• CUDA 9.0+
• mmcv
• pycocotools (pip install lvis@git+https://github.com/open-mmlab/cocoapi.git#subdirectory=lvis)

Installation

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

Future works

• Parse shapefile
• Show polygons or show polygons on corresponding image
• Merge separate polygon in original shapefile
• Parse ignore file (png)
• Add ignore flag to properties
• Show ignored polygons
• Split large image and corresponding polygons
• Convert Json to COCO format
• COCO format visualization codes
• Merge detection results on small image to original image
• Generate CSV file for evaluation (xian fine)
• Evaluation codes for semantic segmentation
• Evaluation codes for instance segmentation
• Visualization code for ground truth CSV file and prediction CSV file
• Visualization code for TP, FN, FP (pred and gt)
• Evaluation codes for offset
• Evaluation codes for height

Structure

• demo: Simple demo to illustrate how to use the corresponding functions
• tools: Put the codes for projects
• bstool
  • datasets: Parse and dump data for dataset (e.g. shapefile, coco, json file)
  • evaluation: Detection and segmentation evaluation codes
  • generation: Generation the specific objects (e.g. empty images, polygons on pixel annotation)
  • ops: Operators (e.g. bbox nms, mask nms)
  • transforms: bbox, mask, image transformation functions (e.g. mask to polygon, polygon to mask)
    • image
    • bbox
    • mask
  • visualization: Codes for visualization
    • image
    • mask
    • bbox
    • color
    • utils
  • utils: Small useful tools for general tasks
    • path
    • mask
  • csrc: Codes for cuda operation (Note: if you meet the compilation errors, please comment this section in setup.py)