Object detection labeling software (auto-labeling)

python version: 3.11.14

Model-assisted object detection labeling tool with FastAPI, Faster R-CNN, and human-in-the-loop correction for dataset creation.

I started this project to quickly label chess pieces using object detection models (auto-labeling).

Currently the label.py script provides a FastAPI server that serves images from the data/ directory and displays the object detection results from the Faster R-CNN model.

This allows manual correction of the object detection labels and save the corrected labels for training purposes, ultimately building a dataset for model fine-tuning.

Installation

pip install -r requirements.txt

Usage

Note: make sure you have a data/ directory with images in it

python label.py

Output

localhost:8000