To train a model run:
cd $PRJ_ROOT/sparsePlane
python tools/train_net.py \
--config-file configs/$CONFIG \
--num-gpus $NUM_GPUwe take three steps to train while freezing the previous part:
| Target | $CONFIG | Pretrain |
|---|---|---|
| Plane | step1-plane.yaml | faster_rcnn_R_50_FPN_3x |
| Embedding | step2-corr.yaml | step1_model_0036999.pth |
| Camera | step3-camera.yaml | step2_model_0035999.pth |