Monodepth2 - UPB

This is the reference PyTorch implementation for training and testing depth estimation models using the method described in

Digging into Self-Supervised Monocular Depth Prediction

Clément Godard, Oisin Mac Aodha, Michael Firman and Gabriel J. Brostow

ICCV 2019

@article{monodepth2,
  title     = {Digging into Self-Supervised Monocular Depth Prediction},
  author    = {Cl{\'{e}}ment Godard and
               Oisin {Mac Aodha} and
               Michael Firman and
               Gabriel J. Brostow},
  booktitle = {The International Conference on Computer Vision (ICCV)},
  month = {October},
year = {2019}
}

Create dataset

mkdir raw_dataset

• Download the UBP dataset into the "raw_dataset" directory. A sample of the UPB dataset is available here. Those video are 3FPS. Consider downloding the original dataset and downsample to 10FPS.

mkdir scene_splits

• Download the scene splits into the "scene_splits" directory. The train-validation split is available here. In the "scene_splits" directory you should have: "train_scenes.txt" and "test_scenes.txt".

# script to create the dataset
python3 scripts/create_dataset.py \
  --src_dir raw_dataset \
  --dst_dir ./dataset \
  --split_dir scene_splits

Train model - example

• Downloading the pretrained model to fine-tune

# script to download pretrained model
python3 download.py

• Fine-tune existing model

# script to train the model
python3 train.py \
  --model_name finetuned_mono \
  --load_weights_folder ./models/mono_640x192 \
  --data_path ./dataset\
  --log_dir ./logs \
  --height 256 \
  --width 512 \
  --num_workers 4 \
  --split upb \
  --dataset upb \
  --learning_rate 1e-6 \
  --batch_size 12 \
  --num_epochs 5 \
  --disparity_smoothness 1e-3\

Conisder playing with "disparity_smoothness".

Test model

• Copy trained model

cp -r logs/finetuned_mono/models/weights_4 models/monodepth

• Get samples

# script to get some sample results
python3 scripts/results.py \
  --model_name monodepth\
  --models_dir ./models\
  --split_dir ./splits/upb\
  --dataset_dir ./dataset\
  --results_dir ./results

Pre-trained model

A pre-trained model (512x256 - 10FPS) is available here.