DenseDepth-Torch

A simple PyTorch Implementation of the "High Quality Monocular Depth Estimation via Transfer Learning" paper.

The paper can be read here.

Official Implementation can be found here.

Training

• The model was trained on the ESPADA dataset.

$ python train.py --epochs 31 --batch 4 --save ./checkpoints/ --device cuda

• The model was trained on Google Colab 30 epochs (~ 7/8 hours), it was trained periodically when Nvidia T4 or P100s were available. Training on a single 12 GB Tesla K80 takes too long. In contrast, the authors use a cluster of 4 Tesla K80s. In contrast, the authors train for 1 M epochs for 20 Hours.

• Train Loss at the end of the 20th epoch was ~0.082.

Usage

• Step 1: Clone the repository

git clone https://github.com/AldrichCabrera/DenseDepth-Torch.git

• Step 2: Download ESPADA dataset or use your own dataset

• Step 3: To train,

python train.py --epochs 31 --batch 4 --save ./checkpoints/ --device cuda

• Step 4: To test,

python test.py --checkpoint ./checkpoints/ckpt_5_7.pth --device cuda --data ./examples/

For help,

python train.py --help