Bbox_handler_config for FasterRCNN?

[nasserha]

I am trying to see how to customize the bounding boxes appearance in inferred images in the case of a fasterrcnn inference.
I saw that detectnetv2 has a bbox_handler_config part where one can control many things. It does not work for fasterrcnn. Is there something equivalent?

Namely, I want to make the boxes look red and ticker and put the label name around.

Thanks

[Morganh]

@nasserha

I want to make the boxes look red and ticker and put the label name around.

You can modify https://github.com/NVIDIA/tao_tensorflow1_backend/blob/main/nvidia_tao_tf1/cv/faster_rcnn/scripts/inference.py.
Refer to
https://github.com/NVIDIA/tao_tensorflow1_backend/blob/main/nvidia_tao_tf1/cv/common/inferencer/inferencer.py#L159-L168

[nasserha]

Hello Morgan,

Where can I find this code?

(launcher) root@machine:~/.virtualenvs/launcher/lib/python3.8/site-packages# ls nvidia*
nvidia_pyindex:
__init__.py  __pycache__  cmdline.py  utils.py

nvidia_pyindex-1.0.9.dist-info:
INSTALLER   METADATA  REQUESTED  entry_points.txt
LICENSE.md  RECORD    WHEEL      top_level.txt

nvidia_tao-5.0.0.dist-info:
INSTALLER  METADATA  REQUESTED  entry_points.txt
LICENSE    RECORD    WHEEL      top_level.txt

nvidia_tao_cli:
__init__.py  __pycache__  components  config  entrypoint  version.py

I went into the installation directory of the environment and looked it up in all folders. I may be in the wrong place :)

[Morganh]

You can login the docker via docker run <5.0 docker> or use tao launcher tao model faster_rcnn .

Then, search the code under /usr