Welcome to MMTracking’s documentation!

You can switch between Chinese and English documents in the lower-left corner of the layout.

Get Started

• Prerequisites
• Installation
  • Detailed Instructions
  • A from-scratch setup script
  • Developing with multiple MMTracking versions
• Verification
• Model Zoo Statistics
• Benchmark and Model Zoo
  • Common settings
  • Baselines of video object detection
  • Baselines of multiple object tracking
  • Baselines of single object tracking
  • Baselines of video instance segmentation

Quick run

• Dataset Preparation
  • 1. Download Datasets
  • 2. Convert Annotations
• Run with Existing Datasets and Models
  • Inference
  • Testing
  • Training
• Run with Customized Datasets and Models
  • 1. Prepare the customized dataset
  • 2. Prepare the customized model
  • 3. Prepare a config
  • 4. Train a new model
  • 5. Test and inference the new model

Tutorials

• Learn about Configs
  • Modify config through script arguments
  • Config File Structure
  • Config Name Style
  • Detailed analysis of Config File
  • FAQ
• Customize Datasets
  • Convert the dataset into CocoVID style
  • Using dataset wrappers
  • Subset of existing datasets
• Customize Data Pipelines
  • Data pipeline for a single image
  • Data pipeline for multiple images
• Customize VID Models
  • Add a new detector
  • Add a new motion model
  • Add a new aggregator
• Customize MOT Models
  • Add a new tracker
  • Add a new detector
  • Add a new motion model
  • Add a new reid model
  • Add a new track head
  • Add a new loss
• Customize SOT Models
  • Add a new backbone
  • Add a new neck
  • Add a new head
  • Add a new loss
• Customize Runtime Settings
  • Customize optimization settings
  • Customize training schedules
  • Customize workflow
  • Customize hooks

Useful Tools and Scripts

• MOT Test-time Parameter Search
• SiameseRPN++ Test-time Parameter Search
• Log Analysis
• Model Conversion
  • Prepare a model for publishing
• Miscellaneous
  • Print the entire config
• Model Serving
  • 1. Convert model from MMTracking to TorchServe
  • 2. Build mmtrack-serve docker image
  • 3. Run mmtrack-serve
  • 4. Test deployment

Notes

• Changelog
  • v0.14.0 (19/09/2022)
  • v0.13.0 (29/04/2022)
  • v0.12.0 (01/04/2022)
  • v0.11.0 (04/03/2022)
  • v0.10.0 (10/02/2022)
  • v0.9.0 (05/01/2022)
  • v0.8.0 (03/10/2021)
  • v0.7.0 (03/09/2021)
  • v0.6.0 (30/07/2021)
  • v0.5.3 (01/07/2021)
  • v0.5.2 (03/06/2021)
  • v0.5.1 (01/02/2021)
  • v0.5.0 (04/01/2021)

Switch Language

• English
• 简体中文

API Reference

• mmtrack.apis
• mmtrack.core
  • anchor
  • evaluation
  • motion
  • optimizer
  • track
  • utils
• mmtrack.datasets
  • datasets
  • parsers
  • pipelines
  • samplers
• mmtrack.models
  • mot
  • sot
  • vid
  • aggregators
  • backbones
  • losses
  • motion
  • reid
  • roi_heads
  • track_heads
  • builder
• mmtrack.utils

Indices and tables

• Index

• Search Page