[Feature] Faster Point Cloud Visualizers #534
Conversation
| colorToDepthTransform.setAndInvert(colorImage.getPose()); | ||
| colorToDepthTransform.multiply(depthPose); | ||
| LibGDXTools.toLibGDX(colorToDepthTransform, tempColorTransform, libGDXColorTransform); |
There was a problem hiding this comment.
Fixes a transform issue. Seems to me that we're using a color to depth transform, not depth to color.
There was a problem hiding this comment.
This guy replaces the RDXROS2ColoredPointCloudVisualizer. I'm not 100% certain about the name. What do we think?
There was a problem hiding this comment.
What about removing the RawImage part of the name?
There was a problem hiding this comment.
Yeah, that's what I was thinking too. It uses the RDXRawImagePointCloudVisualizer, but the API doesn't expose any RawImages so I think it makes sense to remove that part of the name. I'll do that
There was a problem hiding this comment.
Ahhh but RDXROS2PointCloudVisualizer already exists. Should I just steal the RDXROS2ColoredPointCloudVisualizer name instead?
| @@ -31,6 +39,30 @@ public ImageMessageDecoder() | |||
| } | |||
| } | |||
|
|
|||
| public RawImage decodeMessage(ImageMessage messageToDecode) | |||
There was a problem hiding this comment.
Not sure why I didn't make a method that returns a RawImage in the first place. Only dilemma I have is whether to use a CPU or GPU mat. I went with CPU since this class is used on the UI side. Should I add a method that uses GPU mat too?
There was a problem hiding this comment.
I don't think GPU is needed
There was a problem hiding this comment.
I guess in the method name you should put CPU
| @@ -101,7 +101,7 @@ void main() | |||
| #ifdef INPUT_COLOR_IMAGE | |||
| else if (u_coloringMethod == COLOR_FROM_IMAGE) | |||
| { | |||
| vec3 colorFramePoint = transformPoint3D(depthFramePoint, u_depthToColorTransform); | |||
| vec3 colorFramePoint = transformPoint3D(depthFramePoint, u_colorToDepthTransform); | |||
There was a problem hiding this comment.
I think you've got it backwards. You're converting the point from depth frame to color frame. The transform that does that should be called depthToColorTransform.
There was a problem hiding this comment.
Hmm that's true. I'm bad at this stuff so I really don't understand what I'm doing here hahaha. I'll change the names back
There was a problem hiding this comment.
I can explain on the whiteboard whenever
This PR introduces an
RDXRawImagePointCloudVisualizerwhich uses theRDXRawImagePointCloudRenderer. This visualizer also matches up color and depth images based on acquisition time to eliminate de-sync between the two.This PR also replaces the old
RDXROS2ColoredPointCloudVisualizerwith a newRDXROS2RawImagePointCloudVisualizer.Some other helpful methods are added to
TimeToolsandImageMessageDecoder.Video of the new visualizer:
2024-12-13.15-36-00.mp4