silence401/unsup_mvsnet

Trianing and testing on DTU datset.

Training

• Download the preprocessed DTU training data and Depths_raw (both from Original MVSNet), and upzip it as the $MVS_TRANING folder.

├── Cameras    
├── Depths
├── Depths_raw   
├── Rectified
├── Cameras                               
             

• In train.sh, set MVS_TRAINING as $MVS_TRANING
• Train CasMVSNet (Multi-GPU training):

export NGPUS=8
export save_results_dir="./checkpoints"
./train.sh $NGPUS $save_results_dir  --ndepths "48,32,8"  --depth_inter_r "4,2,1"   --dlossw "0.5,1.0,2.0"  --batch_size 2 --eval_freq 3

If apex is installed, you can enable sync_bn in training:

export NGPUS=8
export $save_results_dir="./checkpoints"
./train.sh $NGPUS $save_results_dir  --ndepths "48,32,8"  --depth_inter_r "4,2,1"   --dlossw "0.5,1.0,2.0"  --batch_size 2 --eval_freq 3  --using_apex  --sync_bn

Testing and Fusion

• Download the preprocessed test data DTU testing data (from Original MVSNet) and unzip it as the $TESTPATH folder, which should contain one cams folder, one images folder and one pair.txt file.
• In test.sh, set TESTPATH as $TESTPATH.
• Set CKPT_FILE as your checkpoint file, you also can download my pretrained model.
• Test CasMVSNet and Fusion(default is provided by MVSNet-pytorch):

export save_results_dir="./outputs"
./test.sh  $CKPT_FILE --outdir $save_results_dir  --interval_scale 1.06

• We also support Gipuma to fusion(need to install fusibile) . the script is borrowed from MVSNet.

export save_results_dir="./outputs"
./test.sh  $CKPT_FILE --outdir $save_results_dir  --interval_scale 1.06  --filter_method gipuma

Results on DTU

	Acc.	Comp.	Overall.
MVSNet(D=256)	0.396	0.527	0.462
CasMVSNet(D=48,32,8)	0.325	0.385	0.355

Results on Tanks and Temples benchmark

Mean	Family	Francis	Horse	Lighthouse	M60	Panther	Playground	Train
56.42	76.36	58.45	46.20	55.53	56.11	54.02	58.17	46.56

Please refer to leaderboard.

CasMVSNet input from COLMAP SfM

We use a script provided by MVSNet to convert COLMAP SfM result to CasMVSNet input. After recovering SfM result and undistorting all images, COLMAP should generate a dense folder COLMAP/dense/ containing an undistorted image folder COLMAP/dense/images/ and a undistorted camera folder COLMAP/dense/sparse/. Then, you can use the following command to generate the CasMVSNet input and dense point cloud:

export $save_results_dir="outputs/colmap"
python colmap2mvsnet.py --dense_folder COLMAP/dense  --save_folder $save_scene_result/casmvsnet
./test.sh  $CKPT_FILE  --testpath_single_scene $save_results_dir/casmvsnet  --teslist all --outdir $save_results_dir/ply --interval_scale 1.06