# StereoNet

**Repository Path**: Hoyt_Hu/StereoNet

## Basic Information

- **Project Name**: StereoNet
- **Description**: StereoNet: Guided Hierarchical Refinement for Real-Time Edge-Aware Depth prediction model in pytorch. ECCV2018
- **Primary Language**: Python
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-03-12
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

### StereoNet: Guided Hierarchical Refinement for Real-Time Edge-Aware Depth prediction model in pytorch. ECCV2018

### If you want to communicate with me about the StereoNet, please concact me without hesitating. My email: 
### xuanyili.edu@gmail.com  
### my model result

Now, my model's speed can achieve 60-25FPS on 540*960 img with the best result of 1.87 EPE_all with 16X multi model， 1.95 EPE_all with 16X single model 1.32 EPE_all with 8X single model 1.48EPE_all with 8X multi model on sceneflow dataset by end-to-end training. 
the following are the side outputs and the prediction example
#### train example
![train example](https://github.com/meteorshowers/StereoNet/blob/master/doc/iter-21200.jpg)
#### test example(outputs of 16single model and GT)
![test example](https://github.com/meteorshowers/StereoNet/blob/master/doc/iter-70.jpg)

### Citation
* refercence[1]

If you find our work useful in your research, please consider citing:

@inproceedings{khamis2018stereonet,
  title={Stereonet: Guided hierarchical refinement for real-time edge-aware depth prediction},
  author={Khamis, Sameh and Fanello, Sean and Rhemann, Christoph and Kowdle, Adarsh and Valentin, Julien and Izadi, Shahram},
  booktitle={Proceedings of the European Conference on Computer Vision (ECCV), Munich, Germany},
  pages={8--14},
  year={2018}
}

### Introduction 
I implement the real-time  stereo model according to the   StereoNet   model in pytorch.
The speed can reach 30FPS with top performance.
The speed can reach 60FPS with lower performance.

| Method |EPE_all on sceneflow dataset |EPE_all on kitti2012 dataset|EPE_all on kitti2015 dataset|
|:---|:---:|:---:|:---:|
|ours(16X multi)| 1.32|    |   |
| Reference[1]| 1.525 |    |   |

### License

* Our code is released under MIT License (see LICENSE file for details).

### Installaton

* python3.6
* pytorch0.4

### Usage

* run main8Xmulti.py

### Updates

* finetune the performance beating the original paper.

### To do

* optimize the inference speed

### pretrain model

* coming soon.

### Thanks

* Thanks to  <a href="https://github.com/samehkhamis"> Sameh Khamis' help