YoloSpotSeg

1. Software Introduction

This repository presents an enhanced implementation based on the YOLOv8 object detection framework. Key improvements include: - Addition of Upsampling Layer: An additional upsampling layer is integrated into the original Yolov8 architecture to enhance segmentation performance for spot - Image Block Partitioning and Fourier Interpolation Preprocessing: A preprocessing pipeline is proposed, which involves partitioning images into blocks and applying Fourier interpolation to optimize input data quality. - Gaussian Diffusion Label Conversion Interface: An interface is provided to convert spot center point labels into instance segmentation labels using a Gaussian diffusion function, improving the model's ability to capture target details.

2. Installation Guide

Requirements

• Python >= 3.8
• PyTorch >= 1.13.1
• CUDA >= 11.6

Example conda environment setup

```bash conda create -n yolospotseg python==3.8 conda activate yolospotseg

Clone repository with submodules

git clone --recurse-submodules https://github.com/yourusername/your-repo.git cd your-repo

pip install torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu116 pip install -r requirements.txt

Install custom modules

pip install -e . ```

3. Usage Example

Its usage is fully consistent with the Yolo model.

Preprocess

bash python tools/imgSplit/imgSplit.py --pathRoot /path/to/your/dataset --outputDir /path/save/result --splitFactor 4 - pathRoot: The directory path of the dataset to be preprocessed. - outputDir: The output directory path where the preprocessed files will be saved. - splitFactor:The factor determining how an image is divided into splitFactor × splitFactor blocks.

Train & valid

```bash from ultralytics import YOLO

Load a model

model = YOLO("your_config.yaml")

Train the model

trainresults = model.train( data="yourdataset.yaml", epochs=100, imgsz=640, )

Evaluate model performance on the validation set

metrics = model.val() ```

Web-based interactive usage

For real-time model inference using pre-trained weights on image data, researchers can directly utilize our web-based analytical tool through aclsfip

4. License

This project is built upon Ultralytics Yolov8 which is licensed under AGPL-3.0.

Code Derivative Statement

This software constitutes a derivative work of the YOLOv8 framework (AGPL-3.0 licensed). All modifications and extensions are expressly released under identical open-source terms, maintaining full compliance with the original license's copyleft provisions.

Non-commercial Nature

This implementation contains no commercial components or monetization mechanisms. The codebase is strictly maintained for academic research and open scientific collaboration purposes.

5. Contact

For technical support: - Corresponding author: Liu Huan - Email: lh2022@stu.xjtu.edu.cn