Code for the paper
Mesh R-CNN
Georgia Gkioxari, Jitendra Malik, Justin Johnson
ICCV 2019
The implementation of Mesh R-CNN is based on Detectron2 and PyTorch3D. You will first need to install those in order to be able to run Mesh R-CNN.
To install
git clone https://github.com/facebookresearch/meshrcnn.git
cd meshrcnn && pip install -e .
Run Mesh R-CNN on an input image
python demo/demo.py \
--config-file configs/pix3d/meshrcnn_R50_FPN.yaml \
--input /path/to/image \
--output output_demo \
--onlyhighest MODEL.WEIGHTS meshrcnn://meshrcnn_R50.pth
See demo.py for more details.
See INSTRUCTIONS_PIX3D.md for more instructions.
See INSTRUCTIONS_SHAPENET.md for more instructions.
The Mesh R-CNN codebase is released under BSD-3-Clause License