Skip to content

Latest commit

 

History

History
73 lines (59 loc) · 3.43 KB

vKITTI3D.md

File metadata and controls

73 lines (59 loc) · 3.43 KB
Raw

vKITTI3D

Download all point clouds and labels from the vKITTI3D Dataset and place extracted training files to $VKITTI3D_DIR/data/.

Handcrafted Partition

Not available for this dataset, the sparsity of the acquisition renders the handcrafted geometric features useless.

Learned Partition

For the learned partition, run:

python supervized_partition/graph_processing.py --ROOT_PATH $VKITTI3D_DIR --dataset vkitti --voxel_width 0.05 --use_voronoi 1

for FOLD in 1 2 3 4 5 6; do
    python ./supervized_partition/supervized_partition.py --ROOT_PATH $VKITTI3D_DIR --dataset vkitti \
    --epochs 50 --test_nth_epoch 10  --cvfold $FOLD --reg_strength 0.5 --spatial_emb 0.02 --batch_size 15 \
    --global_feat exyrgb --CP_cutoff 10 --odir results_part/vkitti/best; \
 done;

or use our trained weights and the --resume RESUME argument:

python supervized_partition/graph_processing.py --ROOT_PATH $VKITTI3D_DIR --dataset vkitti --voxel_width 0.05 --use_voronoi 1

for FOLD in 1 2 3 4 5 6; do
   python ./supervized_partition/supervized_partition.py --ROOT_PATH $VKITTI3D_DIR --dataset vkitti \
   --epochs -1 --test_nth_epoch 10 --cvfold $FOLD --reg_strength 0.5 --spatial_emb 0.02 --batch_size 15\
   --global_feat exyrgb --CP_cutoff 10 --odir results_partition/vkitti/pretrained --resume RESUME; \
done;

To evaluate the quality of the partition, run:

python supervized_partition/evaluate_partition.py --dataset vkitti --cvfold 123456 --folder best

Training

Then, reorganize point clouds into superpoints by:

python learning/vkitti_dataset.py --VKITTI_PATH $VKITTI3D_DIR

To train from scratch, run:

for FOLD in 1 2 3 4 5 6; do \ 
CUDA_VISIBLE_DEVICES=0 python ./learning/main.py --dataset vkitti --VKITTI_PATH $VKITTI3D_DIR --cvfold $FOLD --epochs 100 \ --lr_steps "[40, 50, 60, 70, 80]" --test_nth_epoch 10 --model_config gru_10_1_1_1_0,f_13 --pc_attribs xyzXYZrgb \ 
--ptn_nfeat_stn 9 --batch_size 4 --ptn_minpts 15 --spg_augm_order 3 --spg_augm_hardcutoff 256 \
--ptn_widths "[[64,64,128], [64,32,32]]" --ptn_widths_stn "[[32,64], [32,16]]" --loss_weights sqrt \
--use_val_set 1 --odir results/vkitti/best/cv$FOLD; \
done;\

or use our trained weights with the --resume RESUME argument:

for FOLD in 1 2 3 4 5 6; do \ 
CUDA_VISIBLE_DEVICES=0 python ./learning/main.py --dataset vkitti --VKITTI_PATH $VKITTI3D_DIR --cvfold $FOLD --epochs 100 \ --lr_steps "[40, 50, 60, 70, 80]" --test_nth_epoch 10 --model_config gru_10_1_1_1_0,f_13 --pc_attribs xyzXYZrgb \ 
--ptn_nfeat_stn 9 --batch_size 4 --ptn_minpts 15 --spg_augm_order 3 --spg_augm_hardcutoff 256 \
--ptn_widths "[[64,64,128], [64,32,32]]" --ptn_widths_stn "[[32,64], [32,16]]" --loss_weights sqrt \
--use_val_set 1 --odir results/vkitti/pretrained/cv$FOLD --resume RESUME; \
done;\

Estimate the quality of the semantic segmentation with:

python learning/evaluate.py --dataset vkitti --odir results/vkitti/best --cvfold 123456

Visualization

To visualize the results and intermediary steps (on the subsampled graph), use the visualize function in partition. For example:

python partition/visualize.py --dataset vkitti --ROOT_PATH $VKITTI3D_DIR --res_file 'results/vkitti/cv1/predictions_test' --file_path '01/0001_00000' --output_type ifprs