Attention-Aware Multi-View Stereo

The code of the paper Attention-Aware Multi-View Stereo.

The original paper could be found here (CVPR2020).

---

How to use

Our experiment use the dataset of DTU benchmark

0. Dataset

• DTU train part: gdown --id 1_Nuud3lRGaN_DOkeTNOvzwxYa2z2YRbX

1. Clone the source code

git clone https://github.com/soroushhashemifar/Attention-aware-MVSNet.git

2. Download testing dataset

• DTU test part: gdown --id 1rX0EXlUL4prRxrRu2DgLJv2j7-tpUD4D

3. Train the model

python train.py

4. Generate depth map using our pre-trained model

python eval.py

When finished, you can find depth maps in results folder.