This repository provides a PyTorch implementation of the YOLOv8 object detection model, designed for training on custom datasets. It targets researchers and developers needing a flexible and performant object detection solution, offering features like multi-GPU training and adaptive learning rates.

How It Works

The project implements the YOLOv8 architecture in PyTorch, supporting various training configurations including different learning rate schedulers (step, cosine), optimizers (Adam, SGD), and adaptive learning rate adjustments based on batch size. It also includes features like heatmap generation and Exponential Moving Average (EMA) for improved training stability and performance.

Quick Start & Requirements

• Install: pip install torch==1.2.0 (or torch==1.7.1+ for AMP).
• Prerequisites: PyTorch.
• Data: Requires VOC format datasets. Pre-trained weights and VOC dataset links are provided via Baidu Netdisk.
• Docs: Training, prediction, and evaluation steps are detailed in the README.

Highlighted Details

• Supports YOLOv8 variants: n, s, m, l, x.
• Achieves mAP@0.5:0.95 of 52.9 with yolov8_x on COCO-Val2017 at 640x640 input.
• Includes multi-GPU training and adaptive learning rate adjustments.
• Offers features like heatmap generation and EMA.

Maintenance & Community

The repository is maintained by bubbliiiing. No specific community links (Discord, Slack) or roadmap are provided in the README.

Licensing & Compatibility

The repository's license is not explicitly stated in the README. It appears to be a direct implementation of YOLOv8, which is typically under an AGPL-3.0 license by Ultralytics, implying potential copyleft restrictions for commercial use.

Limitations & Caveats

The project requires PyTorch version 1.2.0, with a recommendation for 1.7.1+ for AMP. Data and pre-trained weights are hosted on Baidu Netdisk, which may pose accessibility issues. The license is not clearly specified, which could impact commercial adoption.