YOLO Detection & Utilities

Lightweight repository of scripts and utilities for YOLO-based object detection, SAHI slicing, dataset creation, visualization, and Label Studio integration.

What's changed

• requirements.txt now includes torch and torchvision. For CUDA-enabled installs, please use the official PyTorch selector to install the correct wheel for your GPU.
• A configuration template is available at scripts/Sahi_detect/config_template.yaml (copy and edit to create your runtime config.yaml).

Installation

1. Create and activate a Python virtual environment:

python3 -m venv yolo_env
# Linux / macOS
source yolo_env/bin/activate
# Windows (PowerShell)
.\yolo_env\Scripts\Activate.ps1

2. Install PyTorch (choose proper wheel for your CUDA version). Example for CUDA via PyTorch's index (adjust to your CUDA):

pip install torch torchvision --index-url https://download.pytorch.org/whl/cu130

3. Install remaining dependencies:

pip install -r requirements.txt

Note: requirements.txt includes torch/torchvision as convenience, but installing PyTorch through the official selector ensures compatibility with your CUDA runtime.

Quick Start

• Run a simple detection (example):

python yolo_detect_v11.py --input path/to/image.jpg --conf 0.5

• Run SAHI detection (example):

python scripts/Sahi_detect/sahi_detection.py --config scripts/Sahi_detect/config_template.yaml

Configuration

Use scripts/Sahi_detect/config_template.yaml as a starting point. Copy it to scripts/Sahi_detect/config.yaml and update the input_dir, output_dir, and model.path values to point to your dataset and model weights.

Files of Interest

• yolo_detect_v11.py — YOLO v11 inference script
• scripts/train/yolo11_train.py — Training script
• scripts/Sahi_detect/ — SAHI detection scripts and configs
• utils/ — utilities for staging images, visualization, dataset creation, Label Studio helpers
• requirements.txt — project dependencies

Troubleshooting

• Verify GPU drivers and CUDA with nvidia-smi and test PyTorch with:

python -c "import torch; print('GPU Found!' if torch.cuda.is_available() else 'CPU Only')"

• If Pillow raises errors, see scripts/Sahi_detect/INSTALL_PILLOW.md.

Contributing

Contributions welcome — open an issue or submit a PR.

---

Last Updated: February 2026