├── README.md
└── figures
├── DTU.jpg
├── T&T.jpg
├── teaser.jpg
└── teaser.png
/README.md:
--------------------------------------------------------------------------------
1 | # SurfaceNet+
2 | - An End-to-end 3D Neural Network for Very Sparse MVS.
3 | * 2020TPAMI [early access link](https://ieeexplore.ieee.org/document/9099504).
4 | * or Arxiv [preprint version](https://www.researchgate.net/publication/341647549_SurfaceNet_An_End-to-end_3D_Neural_Network_for_Very_Sparse_Multi-view_Stereopsis/figures).
5 | - **Key contributions**
6 | 1. Proposed a Sparse-MVS benchmark (under construction)
7 | * Comprehensive evaluation on the datasets: [DTU](http://roboimagedata.compute.dtu.dk/?page_id=36), [Tanks and Temples](https://www.tanksandtemples.org/), etc.
8 | 2. Proposed a **trainable occlusion-aware** view selection scheme for the volumetric MVS method, e.g., [SurfaceNet](https://github.com/mjiUST/SurfaceNet)[5].
9 | 3. Analysed the advantages of the volumetric methods, e.g., [SurfaceNet](https://github.com/mjiUST/SurfaceNet)[5] and SurfaceNet+, on the **Sparse-MVS problem** over the depth-fusion methods, e.g., [Gipuma](https://github.com/kysucix/gipuma) [6], [R-MVSNet](https://github.com/YoYo000/MVSNet)[7], [Point-MVSNet](https://github.com/callmeray/PointMVSNet)[8], and [COLMAP](https://github.com/colmap/colmap)[9].
10 |
11 | # [Sparse-MVS Benchmark](http://sparse-mvs.com)
12 |
13 |
14 | 
15 |
16 |
17 | ## (1) [Sparse-MVS of the DTU dataset](http://sparse-mvs.com/leaderboard.html)
18 |
19 |
20 | 
21 |
22 | **Fig.1**: Illustration of a very sparse MVS setting using only $1/7$ of the camera views, i.e., $\{v_i\}_{i=1,8,15,22,...}$, to recover the model 23 in the DTU dataset [10]. Compared with the state-of-the-art methods, the proposed SurfaceNet+ provides much complete reconstruction, especially around the boarder region captured by very sparse views.
23 |
24 |
25 |
26 | 
27 |
28 | **Fig.2**: Comparison with the existing methods in the DTU Dataset [10] with different sparsely sampling strategy. When Sparsity = 3 and Batchsize = 2, the chosen camera indexes are 1,2 / 4,5 / 7,8 / 10,11 / .... SurfaceNet+ constantly outperforms the state-of-the-art methods at all the settings, especially at the very sparse scenario.
29 |
30 |
31 | ## (2) [Sparse-MVS of the T&T dataset](http://sparse-mvs.com/leaderboard.html)
32 |
33 |
34 | 
35 |
36 | **Fig.3**: Results of a tank model in the Tanks and Temples 'intermediate' set [23] compared with R-MVSNet [7] and COLMAP [9], which demonstrate the power of SurfaceNet+ of high recall prediction in the sparse-MVS setting.
37 |
38 |
39 |
40 |
41 | # Citing
42 |
43 | If you find SurfaceNet+, the Sparse-MVS benchmark, or [SurfaceNet](https://github.com/mjiUST/SurfaceNet) useful in your research, please consider citing:
44 |
45 | @article{ji2020surfacenet_plus,
46 | title={SurfaceNet+: An End-to-end 3D Neural Network for Very Sparse Multi-view Stereopsis},
47 | author={Ji, Mengqi and Zhang, Jinzhi and Dai, Qionghai and Fang, Lu},
48 | journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
49 | year={2020},
50 | publisher={IEEE}
51 | }
52 |
53 | @inproceedings{ji2017surfacenet,
54 | title={SurfaceNet: An End-To-End 3D Neural Network for Multiview Stereopsis},
55 | author={Ji, Mengqi and Gall, Juergen and Zheng, Haitian and Liu, Yebin and Fang, Lu},
56 | booktitle={Proceedings of the IEEE International Conference on Computer Vision (ICCV)},
57 | pages={2307--2315},
58 | year={2017}
59 | }
60 |
61 |
--------------------------------------------------------------------------------
/figures/DTU.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/mjiUST/SurfaceNet-plus/38a486000ff66b8b1dda89c04812d77d14023823/figures/DTU.jpg
--------------------------------------------------------------------------------
/figures/T&T.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/mjiUST/SurfaceNet-plus/38a486000ff66b8b1dda89c04812d77d14023823/figures/T&T.jpg
--------------------------------------------------------------------------------
/figures/teaser.jpg:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/mjiUST/SurfaceNet-plus/38a486000ff66b8b1dda89c04812d77d14023823/figures/teaser.jpg
--------------------------------------------------------------------------------
/figures/teaser.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/mjiUST/SurfaceNet-plus/38a486000ff66b8b1dda89c04812d77d14023823/figures/teaser.png
--------------------------------------------------------------------------------