[Fix] enable visualization of the demo results online

[wHao-Wu]

• Enabling visualization of the pcd demo results online
• Enabling visualization of the multi-modality demo results online

Risk: Whether to simultaneously visualize the point and image results online when visualizing multi-modality demo?

[wHao-Wu]

PCD demo

SECOND:

VoteNet:

Multi-modality demo

MVX:

ImVoteNet:

[codecov]

Codecov Report

Merging #473 (19f20f8) into master (6942cf9) will decrease coverage by 0.00%.
The diff coverage is 63.63%.

❗ Current head 19f20f8 differs from pull request most recent head 5a046af. Consider uploading reports for the commit 5a046af to get more accurate results

@@            Coverage Diff             @@
##           master     #473      +/-   ##
==========================================
- Coverage   51.05%   51.04%   -0.01%     
==========================================
  Files         187      187              
  Lines       13799    13801       +2     
  Branches     2238     2238              
==========================================
  Hits         7045     7045              
- Misses       6277     6279       +2     
  Partials      477      477              

Flag	Coverage Δ
unittests	51.04% <63.63%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmdet3d/core/visualizer/show_result.py	64.03% <55.55%> (-1.15%)	⬇️
mmdet3d/apis/inference.py	56.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6942cf9...5a046af. Read the comment docs.

[Wuziyi616]

@wHao-Wu Are we able to visualize img and points at the same time? As far as I know, we need to press q to exit open3d.vis and mmcv.imshow(). How can you show them simultaneously? I know there is a so-called non-blocking mode in opencv but I haven't tried that before.

BTW, in multi-modality browse_dataset, I first show 3D point results, and then show 2D img results when the user exits 3D window.

[Wuziyi616]

I don't think we should save the online results every time, at least let's pass a parameter to determine whether to save? Like what I've done here?

[wHao-Wu]

Thanks for your advice. I have also considered the problem of mandatory saving online results. I will modify the corresponding operation lately.

[wHao-Wu]

@wHao-Wu Are we able to visualize img and points at the same time? As far as I know, we need to press q to exit open3d.vis and mmcv.imshow(). How can you show them simultaneously? I know there is a so-called non-blocking mode in opencv but I haven't tried that before.

BTW, in multi-modality browse_dataset, I first show 3D point results, and then show 2D img results when the user exits 3D window.

I am sorry I have not explained the problem clearly. I mean that maybe we don't need to visualize the points results every time because we have shown the final results in which the detected boxes have been drawn in images.

[Wuziyi616]

@wHao-Wu Are we able to visualize img and points at the same time? As far as I know, we need to press q to exit open3d.vis and mmcv.imshow(). How can you show them simultaneously? I know there is a so-called non-blocking mode in opencv but I haven't tried that before.
BTW, in multi-modality browse_dataset, I first show 3D point results, and then show 2D img results when the user exits 3D window.

I am sorry I have not explained the problem clearly. I mean that maybe we don't need to visualize the points results every time because we have shown the final results in which the detected boxes have been drawn in images.

Maybe we can ask what @ZwwWayne thinks. In my opinion, both 3d points and 2d image are input data so it would be better if we show both?

[baheytharwat]

@wHao-Wu Are we able to visualize img and points at the same time? As far as I know, we need to press q to exit open3d.vis and mmcv.imshow(). How can you show them simultaneously? I know there is a so-called non-blocking mode in opencv but I haven't tried that before.

BTW, in multi-modality browse_dataset, I first show 3D point results, and then show 2D img results when the user exits 3D window.

@Wuziyi616 , Is there a way to update the open3d visualization window without pressing q each time? I am using show_result_meshlab function from the demo example
I just want to display the detection results of multiple point clouds sequentially without pressing any key