Advanced GIS Project -- Segementing Indoor Point Clouds

mmdetection3d

Docker Environment

1. Edit the original dockerfile

&& pip install --no-cache-dir -r ./requirements.txt

2. Run the dockerfile under the mmdetection3d

docker build -t mmdetection3d docker/ 

3. Build the container

docker run -it --name container1 --gpus all \
    -v {/path/to/your/directory}:/workspace \
    -w /workspace \
     mmdetection3d /bin/bash
mim install "mmdet3d>=1.1.0rc0"
docker stop conatiner1

If has done this before, then run:

docker start container1
docker exec -it container1 /bin/bash

or

docker start -ai container1

Don't forget to stop the container in the end.

4. Interference from pre-trained model

Remeber to check the configs and checkpoint files.

python demo/pcd_seg_demo.py demo/data/scannet/scene0000_00.bin pointnet2_msg_2xb16-cosine-250e_scannet-seg-xyz-only.py ${CHECKPOINT_FILE} [--out-dir ${OUT_DIR}] [--show]

Check the output in workplace/output. Normaly, the checkpoint files are saved in demo/cps.