pose_generate.md

Pose estimate check in

Introduction

This part contains the code for implementing pose estimation, virtual character generation, and background data upload on the Jetson platform.

Environment and hardware

• reComputer mini J4012
• USB NFC card reader
• USB Camera
• Preinstalled Jetpack 6.2
• Python 3.10
• Torch 2.7
• Torchvision 0.22

Directory structure

punch_in/
├── main.py                  # Entry point: punch_system startup + NFC thread
├── process.py               # Core punch workflow (pose detection, stickman render, OSS/api updates)
├── punch_config.py          # Shared configuration (model paths, screen settings, skeleton map)
├── read_nfc.py              # Low-level NFC reader CLI built on vendor DLL/so
├── punch_photos/            # Captured stickman images (user punch snapshots)
├── avatars/                 # Cached user avatars downloaded from API
├── models/                  # Pose model binaries
├── icon/                    # UI icon assets (red/green indicators, default avatar)
├── utils/
│   ├── nfc_utils.py         # Shared NFC helpers: TLV parsing, URI building, read/write wrappers
│   ├── read_nfc_url_simple.py
│   ├── write_nfc.py
│   ├── write_nfc_urls.py
│   ├── server_api_example.py
├── libOURMIFARE.so          # Vendor NFC shared library (Linux arm64)
└── README

Pose Check-in Workflow