Skip to content

Latest commit

 

History

History
61 lines (44 loc) · 2.03 KB

README.md

File metadata and controls

61 lines (44 loc) · 2.03 KB
Raw

Depth Estimation with MetaPrompts

Prepare datasets

  1. Prepare NYUDepthV2 datasets following GLPDepth and BTS.
mkdir nyu_depth_v2
wget http://horatio.cs.nyu.edu/mit/silberman/nyu_depth_v2/nyu_depth_v2_labeled.mat
python extract_official_train_test_set_from_mat.py nyu_depth_v2_labeled.mat splits.mat ./nyu_depth_v2/official_splits/

Download sync.zip provided by the authors of BTS from this url and unzip in ./nyu_depth_v2 folder.

  1. Prepare KITTI datasets following GLPDepth and BTS.

Your datasets directory should be:

│nyu_depth_v2/
├──official_splits/
│  ├── test
│  ├── train
├──sync/

│kitti/
├──data_depth_annotated/
├──raw_data/
├──val_selection_cropped/

Results and Fine-tuned Models

NYUv2 RMSE d1 d2 d3 REL Fine-tuned Model
MetaPrompts 0.223 0.976 0.997 0.999 0.061 Google drive
KITTI RMSE d1 d2 d3 REL Fine-tuned Model
MetaPrompts 1.928 0.981 0.998 1.000 0.047 Google drive

Training

Run the following instuction to train the MetaPrompts-Depth model. We recommend using 8 NVIDIA V100 GPUs to train the model with a total batch size of 24.

Train the MetaPrompts-Depth model with 8 NVIDIA V100 GPUs on NYUv2 dataset:

bash train.sh <LOG_DIR>

Train the MetaPrompts-Depth model with 8 NVIDIA V100 GPUs on KITTI dataset:

bash train_kitti.sh <LOG_DIR>

Evaluation

Command format:

bash test.sh/test_kitti.sh <CHECKPOINT_PATH>