HITNet: Hierarchical Iterative Tile Refinement Network for Real-time Stereo Matching

This is a Pytorch implementations of "HITNet: Hierarchical Iterative Tile Refinement Network for Real-time Stereo Matching".

---

Accuracy

Model	Sceneflow Finalpass, EPE	δ<0.1(%)	δ<1.0(%)	δ<3.0(%)	GMac(G)	Checkpoint
HitNet-XL	0.3762	82.1971	96.2759	98.1472	386.6757	ckpt	converted copy of original tensorflow model
HitNet	0.5486	75.1038	94.5830	97.3138	50.5048	ckpt
StereoNet	0.7566	50.0250	91.0111	96.2597	106.7765	ckpt	8x downsample

Training

1. Compile and install cuda op

   pip install ./ext_op

2. Replace dataset path in preprocess/plane_fitting.py and script/hitnet_sf_finalpass.sh

3. Robust plane fitting

   python preprocess/plane_fitting_sf.py
   

4. Training

   bash script/hitnet_sf_finalpass.sh

Evaluation

1. Replace dataset path in eval.py

2. Evaluation

   python eval.py --model HITNet --ckpt ckpt/{ckpt_name} --data_type SceneFlow --data_root_val {path} --data_list_val lists/sceneflow_test.list

Predict

python predict.py --model HITNet --ckpt ckpt/{ckpt_name} --images {left.png} {right.png} --output {disp.png}

Citation

@article{tankovich2020hitnet,
  title={HITNet: Hierarchical Iterative Tile Refinement Network for Real-time Stereo Matching},
  author={Tankovich, Vladimir and H{\"a}ne, Christian and Fanello, Sean and Zhang, Yinda and Izadi, Shahram and Bouaziz, Sofien},
  journal={arXiv preprint arXiv:2007.12140},
  year={2020}
}

However, if you find this implementation or pre-trained models helpful, please consider to cite:

@misc{hang2021tinyhitnet,
  title={TinyHITNet},
  author={zjjMaiMai},
  howpublished={\url{https://github.com/zjjMaiMai/TinyHITNet}},
  year={2021}
}