Contents

- Depth-Anything2-based generalized depth extraction

- Depth-Anything2 TensorRT test

Structures of Project Folders

- Download TensorRT 8.6.1.6

- Depth_Anythingv2_TensorRT_python

${CODE_ROOT}
            |   |-- run_video.py
            |   |-- run_video_trt.py
            |   |-- images_depth_image_extraction.py
            |   |-- videos_depth_image_extraction.py
            |   |-- Depth_Anythingv2_TensorRT_python
            |   |   |   |-- checkpoints
            |   |   |   |   |   |-- depth_anything_v2_vitl.onnx
            |   |   |   |   |   |-- depth_anything_v2_vitl_fp16.engine
            |   |   |   |   |   |-- ...
            |   |   |   |-- tools
            |   |   |   |   |   |-- onnx2trt.py
            |   |   |   |-- ...
            |   |-- TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0
            |   |   |   |-- TensorRT-8.6.1.6
            |   |   |   |   |   |-- onnx_graphsurgeon
            |   |   |   |   |   |-- ...
            |   |   |   |-- ...
            |   |-- ...

Docker Environments

- Build docker environment

docker build -t depth_anything_trt_env .

- Run docker environment

docker run -it --gpus all --name {container_name} \
--shm-size=64G -p 8845:8845 -e GRANT_SUDO=yes --user root \
-v {root_folder}:/workspace/Depth-Anything \
-w /workspace/Depth-Anything depth_anything_trt_env:latest bash

Depth-Anything2 ONNX & TensorRT Engine

- ONNX Download

- TensorRT Engine Sample Download

• Password: 1234

- TensorRT Setup

• This setup is already installed in depth_anything_trt_env. If you need to install tensorrt in a new environment, run it.
• Need to match the Python version

export LD_LIBRARY_PATH=/path/to/TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0/TensorRT-8.6.1.6/lib:$LD_LIBRARY_PATH
export PATH=$PATH:/path/to/TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0/TensorRT-8.6.1.6/bin
source ~/.bashrc
source /etc/profile

cd TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0/TensorRT-8.6.1.6/python
pip install tensorrt-8.6.1-cp310-none-linux_x86_64.whl

cd TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0/TensorRT-8.6.1.6/graphsurgeon
pip install graphsurgeon-0.4.6-py2.py3-none-any.whl

cd TensorRT-8.6.1.6.Linux.x86_64-gnu.cuda-12.0/TensorRT-8.6.1.6/onnx_graphsurgeon
pip install onnx_graphsurgeon-0.3.12-py2.py3-none-any.whl

- Convert ONNX to TensorRT Engine

cd Depth_Anythingv2_TensorRT_python
python tools/onnx2trt.py \
--mode {fp16|fp32} \
-o {onnx_path} \
--output {trt_engine_path} \

Depth Extraction

- Depth extraction from videos

python videos_depth_image_extraction.py \
--encoder {vitl|vitb|vits} \
--video_path {input_video_path} \
--dep_folder_path {depth_save_path} \
--rgb_dep_folder_path {rgb_depth_save_path} \
--raw_img_folder_path {raw_image_save_path} \
--video_out_dir {video_save_path} \
--save_raw_img \
--save_video

- Depth extraction from images

python images_depth_image_extraction.py \
--encoder {vitl|vitb|vits} \
--image_path {input_folder_path} \
--dep_folder_path {depth_save_path} \
--rgb_dep_folder_path {rgb_depth_save_path}

Inference

- Inference video (Pytorch)

python run_video.py \
--encoder {vitl|vitb|vits} \
--video_path {input_video_path}
--outdir {video_save_path}

- Inference video (TensorRT)

python run_video_trt.py \
--engine {trt_engine_path} \
--video_path {input_video_path}
--outdir {video_save_path}

References

- Paper

- Original Code

- TensorRT

Author

- LinkedIn: https://www.linkedin.com/in/taeyong-kong-016bb2154

- Blog URL: https://blog.naver.com/qbxlvnf11

- Email: qbxlvnf11@google.com, qbxlvnf11@naver.com