This repository contains a machine learning-powered web service designed to detect keypoints on a bedsheet from the aligned Color and Depth images. The system is built with Python and uses the FastAPI framework for the web API.

Quick Start

1. Create python virtual runtime environment

python -m venv .bedsheet-endpoints
source .bedsheet-endpoints/bin/activate

2. Install python packages

pip install -r requirements.txt

3. Start FastAPI service

cd src
nohup python -m uvicorn main:app --host ${HOST_FastAPI} --port 8000 --reload &

Folder Structure

bedsheet-endpoints/
├── scripts/
│   └── app_client.py                        # client script here
|   └── save_align-depth2color.py            # script to save the aligned Color and Depth images 
├── src/
│   └── app/                                 # Main application package
│       ├── __init__.py
│       ├── api/
│       │   └── endpoints.py                 # API routes and endpoints
│       ├── core/
│       │   └── config.py                    # Configuration and settings
│       ├── services/
│       │   └── keypoint_detector.py         # Business logic for detection
|       |   └── realsense_capture.py         # Captures the spatial alignment between a color and a depth frame from an Intel RealSense         
│       ├── models/                          # Model network
│       │   ├── utils.py
│       │   └── yolo_vit.py
│       ├── weights/                         # Directory for model weight files
│       │   ├── keypoint_model_vit_depth.pth 
│       │   └── yolo_finetuned/
│       │        ├── best.pt
│       │        └── yolov8l.pt     
│       └── main.py                          # FastAPI app instantiation & startup
|
└── README.md                                # This file

Commands

• check Jetpack version
  apt show nvidia-jetpack

• upgrade PIP to the latest available version
  pip install --upgrade pip

Python packages for JetPack 6.2

• Google >> "pytorch package for jetpack 6.2"
• Python version MUST be 3.10
• Use python -m pip install $PkgName to ensure install package in venv
• Install PyTorch and Torchvision with CUDA support
  • python -m pip install torch==2.8.0 torchvision==0.23.0 --index-url https://pypi.jetson-ai-lab.io/jp6/cu126

Reference

• SDK Manager
  • Jetson AGX Orin 完整刷機
• Install PyTorch on the Jetson Orin Nano running JetPack 6.2
• Installing PyTorch for Jetson Platform
• Your First Jetson Container
  • Cloud-Native on Jetson
• Jetson Orin + RealSense in 5 minutes
  • jetson-orin-librealsense
• Intel® RealSense™ SDK
• Azure-Kinect-Sensor-SDK
  • Azure Kinect SDK
  • pyk4a
• NoMachine -Jetson Remote Desktop
• Chromium browser installation
  • sudo apt install chromium-browser
  • Fixing Browser on Jetson Orin After Update