Welcome to MMYOLO’s documentation!

You can switch between Chinese and English documents in the top-right corner of the layout.

Get Started

• Overview
  • MMYOLO Introduction
  • User guide for this documentation
• Prerequisites
• Installation
  • Best Practices
  • Verify the installation
  • Using MMYOLO with Docker
  • Troubleshooting
• 15 minutes to get started with MMYOLO object detection
  • Installation
  • Dataset
  • Config
  • Training
  • Testing
  • Feature map visualization
  • EasyDeploy deployment
• 15 minutes to get started with MMYOLO rotated object detection
• 15 minutes to get started with MMYOLO instance segmentation
  • Installation
  • Dataset
  • Config
  • Training
  • Testing
  • Feature map visualization
  • EasyDeploy deployment

Recommended Topics

• Contributing to OpenMMLab
  • Preparation
  • Pull Request Workflow
  • Guidance
  • Code style
  • PR Specs
• Training testing tricks
  • Training tricks
  • Testing trick
• Model design instructions
  • YOLO series model basic class
  • HeadModule
• Algorithm principles and implementation
  • Algorithm principles and implementation with YOLOv5
  • Algorithm principles and implementation with YOLOv8
  • Algorithm principles and implementation with RTMDet
• MMYOLO application examples
  • A benchmark for ionogram real-time object detection based on MMYOLO
• Replace the backbone network
  • Use backbone network implemented in MMYOLO
  • Use backbone network implemented in other OpenMMLab repositories
• Model Complexity Analysis
  • Usage Example 1: Print Flops, Parameters and related information by table
  • Usage Example 2: Print related information by network layers
• Annotation-to-deployment workflow for custom dataset
  • 1. Prepare custom dataset
  • 2. Use the software of labelme to annotate
  • 3. Convert the dataset into COCO format
  • 4. Divide dataset into training set, validation set and test set
  • 5. Create a new config file based on the dataset
  • 6. Visual analysis of datasets
  • 7. Optimize Anchor size
  • 8. Visualization the data processing part of config
  • 9. Train
  • 10. Inference
  • 11. Deployment
  • Appendix
• Visualization
  • Feature map visualization
  • Grad-Based and Grad-Free CAM Visualization
  • Perform inference on large images
• MMDeploy deployment tutorial
  • Basic Deployment Guide
  • YOLOv5 Deployment
  • Deploy using Docker
• EasyDeploy deployment tutorial
  • EasyDeploy Deployment
• Troubleshooting steps for common errors
• MM series repo essential basics
• Dataset preparation and description
  • DOTA Dataset

Common Usage

• Resume training
• Automatic mixed precision（AMP）training
• Multi-scale training and testing
  • Multi-scale training
  • Multi-scale testing
• TTA Related Notes
  • Test Time Augmentation (TTA)
• Plugins
• Freeze layers
  • Freeze the weight of backbone
  • Freeze the weight of neck
• Output prediction results
  • Output into json file
  • Output into pkl file
• Set the random seed
• Module combination
• Use mim to run scripts from other OpenMMLab repositories
  • Log Analysis
• Apply multiple Necks
• Specify specific GPUs during training or inference
• Single and multi-channel application examples
  • Training example on a single-channel image dataset
  • Training example on a multi-channel image dataset

Useful Tools

• Visualize COCO labels
• Visualize Datasets
• Print the whole config
• Visualize dataset analysis
• Optimize anchors size
  • k-means
  • Differential Evolution
  • v5-k-means
• Extracts a subset of COCO
• Hyper-parameter Scheduler Visualization
• Dataset Conversion
• Download Dataset
• Log Analysis
  • Curve plotting
  • Compute the average training speed
• Convert Model
  • YOLOv5
  • YOLOX

Basic Tutorials

• Learn about Configs with YOLOv5
  • Config file content
  • Config file inheritance
  • Modify config through script arguments
  • Config name style
• Mixed image data augmentation update
• Customize Installation
  • CUDA versions
  • Install MMEngine without MIM
  • Install MMCV without MIM
  • Install on CPU-only platforms
  • Install on Google Colab
  • Develop using multiple MMYOLO versions
• Common Warning Notes
  • xxx registry in mmyolo did not set import location
  • save_param_schedulers is true but self.param_schedulers is None
  • The loss_cls will be 0. This is a normal phenomenon.
  • The model and loaded state dict do not match exactly
• Frequently Asked Questions
  • Why do we need to launch MMYOLO?
  • What is the projects folder used for?
  • Why does the performance drop significantly by switching the YOLOv5 backbone to Swin?
  • How to use the components implemented in all MM series repositories?
  • Can pure background pictures be added in MMYOLO for training?
  • Is there a script to calculate the inference FPS in MMYOLO?
  • What is the difference between MMDeploy and EasyDeploy?
  • How to check the AP of every category in COCOMetric?
  • Why doesn’t MMYOLO support the auto-learning rate scaling feature as MMDet?
  • Why is the weight size of my trained model larger than the official one?
  • Why does the RTMDet cost more graphics memory during the training than YOLOv5?
  • Do I need to reinstall MMYOLO after modifying some code?
  • How to use multiple versions of MMYOLO to develop?
  • How to save the best checkpoints during the training?
  • How to train and test with non-square input sizes?

Advanced Tutorials

• MMYOLO cross-library application

Model Zoo

• Model Zoo and Benchmark
  • COCO dataset
  • VOC dataset
  • CrowdHuman dataset
  • DOTA 1.0 dataset

Notes

• Changelog
• Compatibility of MMYOLO
• Conventions
• Code Style

API Reference

• mmyolo.datasets
• mmyolo.engine
• mmyolo.models
• mmyolo.utils

Switch Language

• English
• 简体中文

Indices and tables

• Index

• Search Page