pragnyanramtha/churrio

Cheerios Box Detection Project

This project trains a YOLOv8 model to detect Cheerios boxes in images.

Project Overview

The goal of this project is to train a model to detect Cheerios boxes in a dataset of images. The original approach was to use a 3D pose estimation model, but due to several challenges, the project was pivoted to a 2D object detection model using YOLOv8.

Challenges Faced

The initial approach of using the YOLOv5-6D-Pose model for 3D pose estimation was met with several challenges:

• Dependency Issues: The project had strict dependencies on specific versions of Python and PyTorch, which were not compatible with the available hardware.
• Data Formatting: The data format required by the YOLOv5-6D-Pose model was not well-documented, and several attempts to format the data correctly were unsuccessful.
• CUDA Incompatibility: The version of PyTorch required by the YOLOv5-6D-Pose model was not compatible with the available NVIDIA drivers.

Solution

To overcome these challenges, the project was pivoted to a 2D object detection model using YOLOv8. This approach has several advantages:

• Faster Training: YOLOv8 is a much faster model to train than the YOLOv5-6D-Pose model.
• Less Resource-Intensive: YOLOv8 requires fewer resources to train than the YOLOv5-6D-Pose model.
• Simplified Data Preparation: The data preparation for YOLOv8 is much simpler than for the YOLOv5-6D-Pose model.

How to Run

1. Set up the environment:

   uv venv .venv --python 3.9
   source .venv/bin/activate
   uv pip install -r requirements.txt
   

2. Prepare the data:

   python3 prepare_cheerios_yolov8.py
   

3. Train the model:

   yolo train data=YOLOv8/cheerios_yolov8.yaml model=yolov8n.pt epochs=100 imgsz=640