DeepVisionary

English | 中文

English

Overview

DeepVisionary is a modular, PyTorch-based object detection framework inspired by YOLO, designed for developers and researchers to build, train, and evaluate object detection models. It is ideal for learning, experimentation, and small-scale projects, with support for COCO-format datasets and extensible components.

Features

• Modular Architecture: Separates data loading, model components (Backbone, Neck, Head), and training logic for easy customization.
• Configurable: Hyperparameters and dataset paths defined in config.yaml.
• Evaluation & Visualization: Includes mAP computation and visualization tools for detection results.
• Extensible: Supports custom data augmentation, loss functions, and inference pipelines.

Installation

1. Clone the repository:

   git clone https://github.com/colinwps/DeepVisionary.git
   cd DeepVisionary

2. Install dependencies:

   pip install -r requirements.txt

3. Prepare a COCO-format dataset in data/annotations/ and data/images/.

Project Structure

DeepVisionary/
├── configs/               # Configuration files (e.g., config.yaml)
├── data/                  # Data loading and preprocessing
│   ├── dataset.py
│   ├── transforms.py
│   └── annotations/
├── models/                # Model components
│   ├── yolo.py
│   ├── backbone.py
│   ├── neck.py
│   └── head.py
├── scripts/               # Training and inference scripts
│   ├── train.py
│   └── inference.py
├── utils/                 # Utilities
│   ├── loss.py
│   ├── metrics.py
│   ├── logger.py
│   └── visualization.py
├── requirements.txt       # Project dependencies
└── README.md             # Documentation

Usage

1. Configure: Edit configs/config.yaml to set dataset paths and hyperparameters.
2. Train: Run the training script:

   python scripts/train.py

3. Inference: Run inference on a single image:

   python scripts/inference.py --image path/to/image.jpg --checkpoint path/to/model.pth --output output.jpg

   Update class_names in inference.py to match your dataset.
4. Evaluate: Use utils/metrics.py to compute mAP on validation data.
5. Visualize: Use utils/visualization.py to draw bounding boxes on images.

Requirements

• Python >= 3.8
• PyTorch >= 1.9.0
• Torchvision >= 0.10.0
• NumPy >= 1.19.0
• OpenCV-Python >= 4.5.0
• PyYAML >= 5.4.0

Contributing

We welcome contributions! To contribute:

1. Fork the repository.
2. Create a feature branch (git checkout -b feature/your-feature).
3. Commit changes (git commit -m 'Add your feature').
4. Push to the branch (git push origin feature/your-feature).
5. Open a Pull Request.

License

This project is licensed under the MIT License.

---

中文

项目概述

DeepVisionary 是一个基于 PyTorch 的目标检测框架，灵感来源于 YOLO，专为开发者和研究人员设计，适合构建、训练和评估目标检测模型。该项目适用于学习、实验和小规模项目，支持 COCO 格式数据集，并提供可扩展的组件。

功能

• 模块化架构：数据加载、模型组件（主干网络、Neck、Head）和训练逻辑分离，便于定制。
• 可配置：通过 config.yaml 定义超参数和数据集路径。
• 评估与可视化：支持 mAP 计算和检测结果可视化。
• 可扩展：支持自定义数据增强、损失函数和推理流程。

安装

1. 克隆仓库：

   git clone https://github.com/colinwps/DeepVisionary.git
   cd DeepVisionary

2. 安装依赖：

   pip install -r requirements.txt

3. 准备 COCO 格式数据集，放置于 data/annotations/ 和 data/images/。

项目结构

DeepVisionary/
├── configs/               # 配置文件（如 config.yaml）
├── data/                  # 数据加载与预处理
│   ├── dataset.py
│   ├── transforms.py
│   └── annotations/
├── models/                # 模型组件
│   ├── yolo.py
│   ├── backbone.py
│   ├── neck.py
│   └── head.py
├── scripts/               # 训练与推理脚本
│   ├── train.py
│   └── inference.py
├── utils/                 # 工具
│   ├── loss.py
│   ├── metrics.py
│   ├── logger.py
│   └── visualization.py
├── requirements.txt       # 项目依赖
└── README.md             # 项目文档

使用方法

1. 配置：编辑 configs/config.yaml，设置数据集路径和超参数。
2. 训练：运行训练脚本：

   python scripts/train.py

3. 推理：对单张图片进行推理：

   python scripts/inference.py --image path/to/image.jpg --checkpoint path/to/model.pth --output output.jpg

   在 inference.py 中更新 class_names 以匹配你的数据集。
4. 评估：使用 utils/metrics.py 计算验证集的 mAP。
5. 可视化：使用 utils/visualization.py 在图片上绘制边界框。

依赖

• Python >= 3.8
• PyTorch >= 1.9.0
• Torchvision >= 0.10.0
• NumPy >= 1.19.0
• OpenCV-Python >= 4.5.0
• PyYAML >= 5.4.0

贡献

欢迎贡献代码！请按以下步骤操作：

1. Fork 本仓库。
2. 创建特性分支（git checkout -b feature/your-feature）。
3. 提交更改（git commit -m 'Add your feature'）。
4. 推送分支（git push origin feature/your-feature）。
5. 提交 Pull Request。

许可证

本项目采用 MIT 许可证。