annsonic / identify_box_error

Identifying Object Detection Errors for YOLO model

identify_box_error

Identifying Object Detection Errors for YOLO model

Motivation

Inspired by the paper,

@inproceedings{tide-eccv2020,
  author    = {Daniel Bolya and Sean Foley and James Hays and Judy Hoffman},
  title     = {TIDE: A General Toolbox for Identifying Object Detection Errors},
  booktitle = {ECCV},
  year      = {2020},
}

And this reference https://medium.com/data-science-at-microsoft/error-analysis-for-object-detection-models-338cb6534051

Features

• Support YOLO txt format
• Support OBB(oriented bounding box) format defined by YOLOv8
• A summary plot for your model’s errors

  • Error Proportion

    

  • Delta-mAP

    

  • Histogram

    

  • Error annotation

    

Documentation

Install

Create your conda environment and install the dependencies.

$ conda create --name box_analyze python=3.12
$ conda activate box_analyze
$ pip install poetry
$ cd <this_repo>
$ poetry install

Verify the installation

Unit test

$ python -m pytest test

Usage

python main.py 
--yaml_path <the data.yaml used by yolov8> 
--data_subset <train, val or test> 
--predict_folder_path <the output folder used by yolov8>
--analysis_folder_path <the output folder used by this project>
--is_obb

• Ex.

  yaml_path = "datasets/cfg/coco128.yaml"

  data_subset = "tarin"

  predict_folder_path = "/run/detect/"

  analysis_folder_path = "predictions/analyzed_train"

  is_obb is for telling this project to handle an oriented bounding box dataset

Limitation

• Currently, this project reads the cache file generated by yolov8.

  The path of dataset is defined in the yaml file you provide.