This repository contains a collection of Python scripts designed to facilitate the training, testing, and evaluation of object detection models using Detectron2, along with utilities for handling datasets. The scripts cover a range of functionalities from dataset conversion to performance logging.
convertyolotococo.py: Converts YOLO format datasets to COCO format to make them compatible with Detectron2.GOSLAMDetectron_TensorBoard.py: Extends Detectron2's capabilities to include TensorBoard for training visualization.GOSLAMDetectron_Weights_Biases.py: Integrates Detectron2 with Weights & Biases for advanced training metrics tracking and visualization.mergedatasets.py: Combines multiple datasets into a single dataset, facilitating broader training scenarios.test_detectron.py: Provides functionality to test a trained Detectron2 model on new data, outputting performance metrics and visualizations.