Prerequisites

Hardware

• Jetson Orin NX 8GB or higher
• Intel Realsense D435i

Software

• JetPack 5.1.1
• OpenCV 4.10.0-dev with CUDA: YES
  • installation guide: https://qengineering.eu/install-opencv-on-jetson-nano.html
• CUDA 11.4.315
• cuDNN 8.6.0.166
• TensorRT 8.5.2.2
• librealsense 2.55.1.0
  • installation guide: https://github.com/IntelRealSense/librealsense/blob/master/doc/installation_jetson.md
  • command to check libreasense version: sudo dpkg -l | grep librealsense or realsense-viewer --version

Building the Project

Cloning the Repository

git clone https://github.com/kyrie2to11/Depth-Anything-for-Jetson-Orin-CPP.git

Build the Tensor RT engine

cd ./engine
trtexec --onnx=depth_anything_vits14_364.onnx --saveEngine=depth_anything_vits14_364.engine --fp16 

Build executable using CMake

# switch to the project directory
cd xxx/Depth-Anything-for-Jetson-Orin-CPP

# create build directory：./build
cmake -S . -B build

# build the project
cmake --build build

Running the Demo

# usage: 
./depth_anything_cpp [model_trt_engine_path] [--stream or --image or --video] [camera_id (optional:0) or image_path or video path]

# examples:
# 1. run with picture input
./depth_anything_cpp ./engine/depth_anything_vits14_364.engine --image ./test_video_picture/bus.jpg

# 2. run with video input
./depth_anything_cpp ./engine/depth_anything_vits14_364.engine --video ./test_video_picture/davis_dolphins.mp4

# 3. run with live realsense camera input
./depth_anything_cpp ./engine/depth_anything_vits14_364.engine --stream 0

Result Shortcuts

1. run with picture input

2. run with video input:

3. run with live realsense camera input: