MonoDepth

Monocular Depth Estimation

RMS: 5.887 [Baseline: 6.081]

Try it on an image !

1. Download and Unzip Pretrained Model
   • WITH Guided Filtering Layer [Google Drive|BaiduYunPan]
   • WITHOUT Guided Filtering Layer [Google Drive|BaiduYunPan]
2. Run on an Image !
   • WITHOUT Guided Filtering Layer

   python monodepth_simple.py --image_path [IMAGE_PATH] --checkpoint_path [MODEL_PATH]

   • WITH Guided Filter as PostProcessing

   python monodepth_simple.py --image_path [IMAGE_PATH] --checkpoint_path [MODEL_PATH] --guided_filter_eval

   • WITH Guided Filtering Layer

   python monodepth_simple.py --image_path [IMAGE_PATH] --checkpoint_path [MODEL_PATH] --guided_filter

Training on KITTI

1. Download KITTI

   wget -i utils/kitti_archives_to_download.txt -P [SAVE_FOLDER]

2. (Optional) Convert *.png to *.jpg to save space.

   find [SAVE_FOLDER] -name '*.png' | parallel 'convert {.}.png {.}.jpg && rm {}'

3. Let's train the model !
   • WITHOUT Guided Filtering Layer

     python monodepth_main.py --mode train \
                              --model_name monodepth_kitti \
                              --data_path [SAVE_FOLDER] \
                              --filenames_file utils/filenames/kitti_train_files.txt \
                              --log_directory checkpoints

   • WITH Guided Filtering Layer

     python monodepth_main.py --mode train \
                              --model_name monodepth_kitti_dgf \
                              --data_path [SAVE_FOLDER] \
                              --filenames_file utils/filenames/kitti_train_files.txt \
                              --log_directory checkpoints \
                              --guided_filter

4. Download Test Dataset
5. Testing
   • WITHOUT Guided Filtering Layer

     python monodepth_main.py --mode test \
                              --data_path [SAVE_FOLDER] \
                              --filenames_file utils/filenames/kitti_stereo_2015_test_files.txt \
                              --checkpoint_path [MODEL_PATH]

   • WITH Guided Filter as PostProcessing

     python monodepth_main.py --mode test \
                              --data_path [SAVE_FOLDER] \
                              --filenames_file utils/filenames/kitti_stereo_2015_test_files.txt \
                              --checkpoint_path [MODEL_PATH] \
                              --guided_filter_eval

   • WITH Guided Filtering Layer

     python monodepth_main.py --mode test \
                              --data_path [SAVE_FOLDER] \
                              --filenames_file utils/filenames/kitti_stereo_2015_test_files.txt \
                              --checkpoint_path [MODEL_PATH] \
                              --guided_filter

6. Evaluation on KITTI

   python utils/evaluate_kitti.py --split kitti --predicted_disp_path [disparities.npy] --gt_path [SAVE_FOLDER]

Acknowledgement

A part of the code has been borrowed from monodepth.