Skip to content

Commit 7236e08

Browse files
wangxiaomeng030wangxiaomeng030
committed
sync with paper
1 parent f0366f2 commit 7236e08

File tree

3 files changed

+62
-44
lines changed

3 files changed

+62
-44
lines changed

‎README.md

Lines changed: 31 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -23,9 +23,27 @@ We provide a set of pre-implemented deep-learning based SLAM algorithms.
2323

2424
**Replica/office0**
2525

26-
| [nice-slam](https://github.com/cvg/nice-slam) | [co-slam](https://github.com/HengyiWang/Co-SLAM) | [Vox-Fusion](https://github.com/zju3dv/Vox-Fusion) | [Point_SLAM](https://github.com/eriksandstroem/Point-SLAM) | [splaTAM](https://github.com/spla-tam/SplaTAM) | [DPVO](https://github.com/princeton-vl/DPVO) | [NeuralRecon](https://github.com/zju3dv/NeuralRecon) |
27-
| :----------------------------------------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ | ------------------------------------------------------------ |
28-
| <img src="./docs/imgs/nice_slam.gif" alt="nice-slam" style="zoom: 50%;" /> | <img src="./docs/imgs/co_slam.gif" alt="nice-slam" style="zoom: 50%;" /> | <img src="./docs/imgs/vox_fusion.gif" alt="nice-slam" style="zoom: 50%;" /> | <img src="./docs/imgs/point_slam.gif" alt="nice-slam" style="zoom: 50%;" /> | <img src="./docs/imgs/splatam.gif" alt="nice-slam" style="zoom: 50%;" /> | <img src="./docs/imgs/dpvo.gif" alt="dpvo" style="zoom: 50%;" /> | <img src="./docs/imgs/neural_recon.gif" alt="neural_recon" style="zoom: 50%;" /> |
26+
<table>
27+
<tr>
28+
<th style="width: 150px;"><a href="https://github.com/cvg/nice-slam">NICE-SLAM</a></th>
29+
<th style="width: 150px;"><a href="https://github.com/HengyiWang/Co-SLAM">Co-SLAM</a></th>
30+
<th style="width: 150px;"><a href="https://github.com/zju3dv/Vox-Fusion">Vox-Fusion</a></th>
31+
<th style="width: 150px;"><a href="https://github.com/eriksandstroem/Point-SLAM">Point-SLAM</a></th>
32+
<th style="width: 150px;"><a href="https://github.com/spla-tam/SplaTAM">SplaTAM</a></th>
33+
<th style="width: 150px;"><a href="https://github.com/princeton-vl/DPVO">DPVO</a></th>
34+
<th style="width: 150px;"><a href="https://github.com/zju3dv/NeuralRecon">NeuralRecon</a></th>
35+
</tr>
36+
<tr>
37+
<td><img src="./docs/imgs/nice_slam.gif" alt="nice-slam" style="zoom: 50%;" /></td>
38+
<td><img src="./docs/imgs/co_slam.gif" alt="co-slam" style="zoom: 50%;" /></td>
39+
<td><img src="./docs/imgs/vox_fusion.gif" alt="vox-fusion" style="zoom: 50%;" /></td>
40+
<td><img src="./docs/imgs/point_slam.gif" alt="point-slam" style="zoom: 50%;" /></td>
41+
<td><img src="./docs/imgs/splatam.gif" alt="splaTAM" style="zoom: 50%;" /></td>
42+
<td><img src="./docs/imgs/dpvo.gif" alt="dpvo" style="zoom: 50%;" /></td>
43+
<td><img src="./docs/imgs/neural_recon.gif" alt="neural_recon" style="zoom: 50%;" /></td>
44+
</tr>
45+
</table>
46+
2947

3048
## Quickstart
3149

@@ -178,7 +196,7 @@ The figure below is the algorithm pipeline. When adding a new deep-learning base
178196

179197
Here are the comparison results on **Replica** datasets. The results of the original algorithm comes from multiple papers.
180198

181-
The algorithms with _X suffix are the corresponding algorithms in the XRDSLAM framework. For details, see [benchmark](docs/benchmark.md).
199+
The algorithms with * suffix are the corresponding algorithms in the XRDSLAM framework. For details, see [benchmark](docs/benchmark.md).
182200

183201
(DPVO is a visual odometry (VO) algorithm that does not generate meshes, and DPVO paper does not provide results for the Replica datasets. Therefore, here we only present the trajectory metrics from XRDSLAM framework. The comparison results of DPVO for the Euroc dataset can be found in the [benchmark](docs/benchmark.md). NeuralRecon is a mapping algorithm, and NeuralRecon paper does not provide results for the Replica datasets. Therefore, here we only present the mesh metrics from XRDSLAM framework. The comparison results of NeuralRecon for the 7Scenes dataset can be found in the [benchmark](docs/benchmark.md).)
184202

@@ -187,17 +205,17 @@ Note: The default configuration in the algorithm is suitable for Replica. If you
187205
| Algorithm | ATE RMSE [cm] - | PSNR+ | SSIM+ | LPIPS- | Precision [%] + | Recall [%] + | F1[%] + | Depth L1[cm] - | Acc. [cm]- | Comp. [cm]- | Comp. Ratio [<5cm %] + |
188206
| ------------ | ------------------- | ----- | ----- | ------ | --------------- | ------------ | ------- | -------------- | -------------- | -------------- | ---------------------- |
189207
| NICE-SLAM | 1.95 | 24.42 | 0.81 | 0.23 | 44.10 | 43.69 | 43.86 | 3.53 | 2.85 | 3.00 | 89.33 |
190-
| NICE-SLAM_X | 2.09 | 25.68 | 0.85 | 0.32 | 46.62 | 37.53 | 41.47 | 2.62 | 2.03 | 3.38 | 87.81 |
191-
| Co-SLAM | 1.00 | 30.24 | 0.93 | 0.25 | - | - | - | 1.51 | 2.10 | 2.08 | 93.44 |
192-
| Co-SLAM_X | 1.11 | 30.34 | 0.93 | 0.24 | 80.66 | 68.79 | 74.23 | 1.63 | 1.53 | 2.90 | 89.81 |
208+
| NICE-SLAM* | 2.09 | 25.68 | 0.85 | 0.32 | 46.62 | 37.53 | 41.47 | 2.62 | 2.03 | 3.38 | 87.81 |
209+
| Co-SLAM | 0.86 | 30.24 | 0.93 | 0.25 | - | - | - | 1.51 | 2.10 | 2.08 | 93.44 |
210+
| Co-SLAM* | 1.11 | 30.34 | 0.93 | 0.24 | 80.66 | 68.79 | 74.23 | 1.63 | 1.53 | 2.90 | 89.81 |
193211
| Vox-Fusion | 0.54 | 24.41 | 0.80 | 0.24 | 55.73 | 49.13 | 52.20 | 2.46 | 2.37 | 2.28 | 92.86 |
194-
| Vox-Fusion_X | 0.56 | 27.95 | 0.90 | 0.25 | 89.52 | 71.34 | 79.39 | 1.03 | 1.39 | 2.82 | 90.13 |
212+
| Vox-Fusion* | 0.56 | 27.95 | 0.90 | 0.25 | 89.52 | 71.34 | 79.39 | 1.03 | 1.39 | 2.82 | 90.13 |
195213
| Point-SLAM | 0.52 | 35.17 | 0.97 | 0.12 | 96.99 | 83.59 | 89.77 | 0.44 | - | - | - |
196-
| Point-SLAM_X | 0.47 | 34.10 | 0.97 | 0.10 | 99.30 | 83.78 | 90.86 | 0.38 | 1.25 | 3.12 | 88.15 |
214+
| Point-SLAM* | 0.47 | 34.10 | 0.97 | 0.10 | 99.30 | 83.78 | 90.86 | 0.38 | 1.25 | 3.12 | 88.15 |
197215
| SplaTAM | 0.36 | 34.11 | 0.97 | 0.10 | - | - | - | - | - | - | - |
198-
| SplaTAM_X | 0.40 | 34.44 | 0.96 | 0.09 | - | - | - | - | - | - | - |
199-
| DPVO_X | 0.31 | - | - | - | - | - | - | - | - | - | - |
200-
| NeuralRecon_X | - | - | - | - | 13.29 | 7.43 | 9.51 | - | 5.87 | 19.36 | 38.13 |
216+
| SplaTAM* | 0.40 | 34.44 | 0.96 | 0.09 | - | - | - | - | - | - | - |
217+
| DPVO* | 0.31 | - | - | - | - | - | - | - | - | - | - |
218+
| NeuralRecon* | - | - | - | - | 13.29 | 7.43 | 9.51 | - | 5.87 | 19.36 | 38.13 |
201219

202220
## License
203221

@@ -206,7 +224,7 @@ Please note that this license only applies to the code in our library, the depen
206224

207225
## Acknowledgement
208226

209-
In addition to the implemented algorithm ([nice-slam](https://github.com/cvg/nice-slam),[co-slam](https://github.com/HengyiWang/Co-SLAM),[Vox-Fusion](https://github.com/zju3dv/Vox-Fusion),[Point_SLAM](https://github.com/eriksandstroem/Point-SLAM),[splaTAM](https://github.com/spla-tam/SplaTAM), [DPVO](https://github.com/princeton-vl/DPVO), [NeuralRecon](https://github.com/zju3dv/NeuralRecon)), our code also adapt some codes from [nerfStudio](https://github.com/nerfstudio-project/nerfstudio/), [sdfstudio](https://autonomousvision.github.io/sdfstudio/). Thanks for making the code available.
227+
In addition to the implemented algorithm ([NICE-SLAM](https://github.com/cvg/nice-slam),[Co-SLAM](https://github.com/HengyiWang/Co-SLAM),[Vox-Fusion](https://github.com/zju3dv/Vox-Fusion),[Point-SLAM](https://github.com/eriksandstroem/Point-SLAM),[SplaTAM](https://github.com/spla-tam/SplaTAM), [DPVO](https://github.com/princeton-vl/DPVO), [NeuralRecon](https://github.com/zju3dv/NeuralRecon)), our code also adapt some codes from [Nerfstudio](https://github.com/nerfstudio-project/nerfstudio/), [Sdfstudio](https://autonomousvision.github.io/sdfstudio/). Thanks for making the code available.
210228

211229
## Built On
212230

0 commit comments

Comments
 (0)