# open-mmlab/mmpose architecture

> Architecture diagram of the GitHub repository open-mmlab/mmpose, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-01). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/open-mmlab/mmpose
- Repository: https://github.com/open-mmlab/mmpose

## Overview

MMPose is a PyTorch toolbox for 2D and 3D pose estimation across human, animal, face, hand, whole-body, and fashion tasks. A typical inference flow takes an image or video through an inferencer, prepares pose inputs, runs a configured model, decodes keypoints, and returns predictions that can be visualized. The repository also provides dataset and evaluation components and configurable model families. Edges reflect explicit imports in indexed source or documented capabilities; unsampled implementation details are not inferred.

Main entry points: [`mmpose/apis/inferencers/base_mmpose_inferencer.py`](https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/base_mmpose_inferencer.py) (Pose Inferencer), [`mmpose/apis/inferencers/pose3d_inferencer.py`](https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/pose3d_inferencer.py) (3D Inferencer), [`configs`](https://github.com/open-mmlab/mmpose/tree/main/configs) (Task Configurations), [`mmpose/apis/inference.py`](https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inference.py) (Inference API)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_api["Inference APIs"]
  node_inferencer["Pose Inferencer"]
  node_pose2d["2D Inferencer"]
  node_pose3d["3D Inferencer"]
  node_inference["Inference API<br/>[inference.py]"]
  node_inference3d["3D Inference<br/>[inference_3d.py]"]
end

subgraph group_data["Pose Data"]
  node_dataset["Datasets"]
  node_transforms["Data Transforms"]
  node_pose_data["Pose Structures"]
end

subgraph group_models["Model Components"]
  node_configs["Task Configurations"]
  node_registry["Component Registry<br/>[registry.py]"]
  node_backbones["Feature Backbones"]
  node_heads["Pose Heads"]
  node_codecs["Keypoint Codecs"]
  node_ae_codec["Associative Embedding"]
end

subgraph group_results["Evaluation &amp; Output"]
  node_metrics["Keypoint Metrics"]
  node_keypoint_eval["Evaluation Functions<br/>[keypoint_eval.py]"]
  node_visualizer["Pose Visualizer"]
end

subgraph group_apps["Applications"]
  node_rtmpose_app["RTMPose Demo<br/>[app.py]"]
  node_dance_app["Dance Demo<br/>[app.py]"]
  node_video_processor["Video Processor<br/>[process_video.py]"]
  node_torchserve["TorchServe Handler<br/>[mmpose_handler.py]"]
end

node_user(("User"))
node_image_video["Image or Video"]
node_predictions["Pose Predictions"]

node_user -->|"requests inference"| node_inferencer
node_image_video -->|"provides input"| node_inferencer
node_inferencer -->|"calls"| node_inference
node_pose2d -->|"extends"| node_inferencer
node_pose3d -->|"extends"| node_inferencer
node_pose3d -->|"uses"| node_pose2d
node_pose3d -->|"uses"| node_inference3d
node_inference -->|"returns samples"| node_pose_data
node_configs -.->|"configures"| node_dataset
node_configs -.->|"configures"| node_backbones
node_configs -.->|"configures"| node_heads
node_transforms -->|"uses"| node_codecs
node_ae_codec -->|"extends"| node_codecs
node_codecs -->|"registers codecs"| node_registry
node_heads -.->|"registers models"| node_registry
node_backbones -->|"registers models"| node_registry
node_metrics -->|"calls"| node_keypoint_eval
node_inferencer -->|"visualizes"| node_visualizer
node_inferencer -->|"produces"| node_predictions
node_rtmpose_app -->|"calls"| node_inferencer
node_dance_app -->|"calls"| node_video_processor
node_torchserve -->|"uses API"| node_inference

click node_inferencer "https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/base_mmpose_inferencer.py"
click node_pose2d "https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/pose2d_inferencer.py"
click node_pose3d "https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/pose3d_inferencer.py"
click node_inference "https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inference.py"
click node_inference3d "https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inference_3d.py"
click node_dataset "https://github.com/open-mmlab/mmpose/tree/main/mmpose/datasets/datasets"
click node_transforms "https://github.com/open-mmlab/mmpose/blob/main/mmpose/datasets/transforms/common_transforms.py"
click node_pose_data "https://github.com/open-mmlab/mmpose/blob/main/mmpose/structures/pose_data_sample.py"
click node_configs "https://github.com/open-mmlab/mmpose/tree/main/configs"
click node_registry "https://github.com/open-mmlab/mmpose/blob/main/mmpose/registry.py"
click node_backbones "https://github.com/open-mmlab/mmpose/tree/main/mmpose/models/backbones"
click node_heads "https://github.com/open-mmlab/mmpose/tree/main/mmpose/models/heads"
click node_codecs "https://github.com/open-mmlab/mmpose/tree/main/mmpose/codecs"
click node_ae_codec "https://github.com/open-mmlab/mmpose/blob/main/mmpose/codecs/associative_embedding.py"
click node_metrics "https://github.com/open-mmlab/mmpose/blob/main/mmpose/evaluation/metrics/keypoint_2d_metrics.py"
click node_keypoint_eval "https://github.com/open-mmlab/mmpose/blob/main/mmpose/evaluation/functional/keypoint_eval.py"
click node_visualizer "https://github.com/open-mmlab/mmpose/blob/main/mmpose/visualization/local_visualizer.py"
click node_rtmpose_app "https://github.com/open-mmlab/mmpose/blob/main/projects/rtmpose/app.py"
click node_dance_app "https://github.com/open-mmlab/mmpose/blob/main/projects/just_dance/app.py"
click node_video_processor "https://github.com/open-mmlab/mmpose/blob/main/projects/just_dance/process_video.py"
click node_torchserve "https://github.com/open-mmlab/mmpose/blob/main/tools/torchserve/mmpose_handler.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_inferencer,node_pose2d,node_pose3d,node_inference,node_inference3d,node_user toneBlue
class node_dataset,node_transforms,node_pose_data toneAmber
class node_configs,node_registry,node_backbones,node_heads,node_codecs,node_ae_codec toneMint
class node_metrics,node_keypoint_eval,node_visualizer toneRose
class node_rtmpose_app,node_dance_app,node_video_processor,node_torchserve,node_image_video,node_predictions toneIndigo
```

## Components

### Inference APIs

- [Pose Inferencer](https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/base_mmpose_inferencer.py): `mmpose/apis/inferencers/base_mmpose_inferencer.py`
- [2D Inferencer](https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/pose2d_inferencer.py): `mmpose/apis/inferencers/pose2d_inferencer.py`
- [3D Inferencer](https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/pose3d_inferencer.py): `mmpose/apis/inferencers/pose3d_inferencer.py`
- [Inference API](https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inference.py): `mmpose/apis/inference.py`
- [3D Inference](https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inference_3d.py): `mmpose/apis/inference_3d.py`

### Pose Data

- [Datasets](https://github.com/open-mmlab/mmpose/tree/main/mmpose/datasets/datasets): `mmpose/datasets/datasets`
- [Data Transforms](https://github.com/open-mmlab/mmpose/blob/main/mmpose/datasets/transforms/common_transforms.py): `mmpose/datasets/transforms/common_transforms.py`
- [Pose Structures](https://github.com/open-mmlab/mmpose/blob/main/mmpose/structures/pose_data_sample.py): `mmpose/structures/pose_data_sample.py`

### Model Components

- [Task Configurations](https://github.com/open-mmlab/mmpose/tree/main/configs): `configs`
- [Component Registry](https://github.com/open-mmlab/mmpose/blob/main/mmpose/registry.py): `mmpose/registry.py`
- [Feature Backbones](https://github.com/open-mmlab/mmpose/tree/main/mmpose/models/backbones): `mmpose/models/backbones`
- [Pose Heads](https://github.com/open-mmlab/mmpose/tree/main/mmpose/models/heads): `mmpose/models/heads`
- [Keypoint Codecs](https://github.com/open-mmlab/mmpose/tree/main/mmpose/codecs): `mmpose/codecs`
- [Associative Embedding](https://github.com/open-mmlab/mmpose/blob/main/mmpose/codecs/associative_embedding.py): `mmpose/codecs/associative_embedding.py`

### Evaluation & Output

- [Keypoint Metrics](https://github.com/open-mmlab/mmpose/blob/main/mmpose/evaluation/metrics/keypoint_2d_metrics.py): `mmpose/evaluation/metrics/keypoint_2d_metrics.py`
- [Evaluation Functions](https://github.com/open-mmlab/mmpose/blob/main/mmpose/evaluation/functional/keypoint_eval.py): `mmpose/evaluation/functional/keypoint_eval.py`
- [Pose Visualizer](https://github.com/open-mmlab/mmpose/blob/main/mmpose/visualization/local_visualizer.py): `mmpose/visualization/local_visualizer.py`

### Applications

- [RTMPose Demo](https://github.com/open-mmlab/mmpose/blob/main/projects/rtmpose/app.py): `projects/rtmpose/app.py`
- [Dance Demo](https://github.com/open-mmlab/mmpose/blob/main/projects/just_dance/app.py): `projects/just_dance/app.py`
- [Video Processor](https://github.com/open-mmlab/mmpose/blob/main/projects/just_dance/process_video.py): `projects/just_dance/process_video.py`
- [TorchServe Handler](https://github.com/open-mmlab/mmpose/blob/main/tools/torchserve/mmpose_handler.py): `tools/torchserve/mmpose_handler.py`

### Other components

- **User**
- **Image or Video**
- **Pose Predictions**

## Connections

- User → Pose Inferencer: requests inference (evidence: `README.md`)
- Image or Video → Pose Inferencer: provides input (evidence: `mmpose/apis/inferencers/base_mmpose_inferencer.py`)
- Pose Inferencer → Inference API: calls (evidence: `mmpose/apis/inferencers/base_mmpose_inferencer.py`)
- 2D Inferencer → Pose Inferencer: extends (evidence: `mmpose/apis/inferencers/pose2d_inferencer.py`)
- 3D Inferencer → Pose Inferencer: extends (evidence: `mmpose/apis/inferencers/pose3d_inferencer.py`)
- 3D Inferencer → 2D Inferencer: uses (evidence: `mmpose/apis/inferencers/pose3d_inferencer.py`)
- 3D Inferencer → 3D Inference: uses (evidence: `mmpose/apis/inferencers/pose3d_inferencer.py`)
- Inference API → Pose Structures: returns samples (evidence: `mmpose/apis/inference.py`)
- Task Configurations → Datasets: configures (evidence: `README.md`)
- Task Configurations → Feature Backbones: configures (evidence: `README.md`)
- Task Configurations → Pose Heads: configures (evidence: `README.md`)
- Data Transforms → Keypoint Codecs: uses (evidence: `mmpose/datasets/transforms/common_transforms.py`)
- Associative Embedding → Keypoint Codecs: extends (evidence: `mmpose/codecs/associative_embedding.py`)
- Keypoint Codecs → Component Registry: registers codecs (evidence: `mmpose/codecs/associative_embedding.py`)
- Pose Heads → Component Registry: registers models
- Feature Backbones → Component Registry: registers models (evidence: `mmpose/models/backbones/litehrnet.py`)
- Keypoint Metrics → Evaluation Functions: calls (evidence: `mmpose/evaluation/metrics/keypoint_2d_metrics.py`)
- Pose Inferencer → Pose Visualizer: visualizes (evidence: `mmpose/apis/inferencers/base_mmpose_inferencer.py`)
- Pose Inferencer → Pose Predictions: produces (evidence: `mmpose/apis/inferencers/base_mmpose_inferencer.py`)
- RTMPose Demo → Pose Inferencer: calls (evidence: `projects/rtmpose/app.py`)
- Dance Demo → Video Processor: calls (evidence: `projects/just_dance/app.py`)
- TorchServe Handler → Inference API: uses API (evidence: `tools/torchserve/mmpose_handler.py`)
