dontLoveBugs / FCRN_pytorch

Pytorch Implementation of Deeper Depth Prediction with Fully Convolutional Residual Networks

FCRN implemented in Pytorch 0.4.1

Introduction

This is a PyTorch(0.4.1) implementation of Deeper Depth Prediction with Fully Convolutional Residual Networks. It can use Fully Convolutional Residual Networks to realize monocular depth prediction. Currently, we can train FCRN using NYUDepthv2 and Kitti Odometry Dataset.

Result

NYU Depthv2

The code was tested with Python 3.5 with Pytorch 0.4.1 in 12GB TITAN X. We train 60 epochs with batch size = 16. The trained model can be download from BaiduYun.

Method	rml	rmse	log10	Delta1	Delta2	Delta3
FCRN	0.127	0.573	0.055	0.811	0.953	0.988
FCRN_ours	0.149	0.527	0.062	0.805	0.954	0.987

Kitti Odometry

Method	rml	rmse	log10	Delta1	Delta2	Delta3
FCRN_ours	0.113	4.801	0.048	0.865	0.957	0.984

Installation

The code was tested with Python 3.5 with Pytorch 0.4.1 in 2 GPU TITAN X.

0. Clone the repo:

   git clone [e-mail]:dontLoveBugs/FCRN_pyotrch.git
   cd FCRN_pytorch

1. Install dependencies:

   For PyTorch dependency, see pytorch.org for more details.

   For custom dependencies:

   pip install matplotlib pillow tensorboardX

2. Configure your dataset path in "dataloaders/path.py".

3. Training

   To train NYU Depth v2, please do:

   python main.py --dataset nyu

   To train it on KITTI, please do:

   python main.py --dataset kitti