Add MAKE_ORGANIZED parameter #789

karnikram · 2018-06-23T18:49:36Z

PR Description

Added MAKE_ORGANIZED parameter to T3DPointsProjectionParams to be able to generate an organized pcl point cloud from a 2D range image
Appropriately set pcl cloud dimensions
More of an "ad-hoc" solution, just to get pcl::IntegralNormalEstimation to work for my gsoc project. Please let me know if I need to make more changes.

I acknowledge to have:

Read the CONTRIBUTING page

(Notify: @MRPT/owners )

jlblancoc

Good! But please, take a look at my review comments. If possible, please do a git rebase so the changes all appear as one single commit (the mrpt git tree is quite large already...)

jlblancoc · 2018-06-24T06:45:14Z

libs/maps/include/mrpt/maps/CPointsMap.h

@@ -1200,6 +1200,10 @@ class PointCloudAdapter<mrpt::maps::CPointsMap>
 	inline size_t size() const { return m_obj.size(); }
 	/** Set number of points (to uninitialized values) */
 	inline void resize(const size_t N) { m_obj.resize(N); }
+	/** Does nothing as of now */
+	inline void setDimensions(const size_t& height, const size_t& width)


I think this method does NOT need to be added here... it's enough to have the empty setDimensions() method in the PointCloudAdapter specialization for all MRPT classes, and only make it to have a real effect in the PCL specialization, right?

I had to add the method to other specializations too because project3DPointsFromDepthImageInto gets instantiated using the non-pcl types too. For instance, here.

because project3DPointsFromDepthImageInto gets instantiated using the non-pcl types too.

Sure... but if I recall it right, it internally instantiates a PointCloudAdapter (PCA) of the required target type as template argument. Then, my point is: shouldn't it be setDimensions() a method of the PCA instead of the base, target types? In this way, only one PCA should be modified to add a dummy (empty) setDimensions() method, the version for all MRPT classes. Only the specialization for PCL should have a setDimensions() method with real stuff inside.

Let me know if this makes now sense to you...

Yes, that makes sense. I did try to do that at first, but I couldn't actually find the definition for the class template PointCloudAdapter adapter anywhere... The class declaration is here, but only specialization specific definitions have been written, I think.

Or maybe I'm wrong..

Should I try to write the base template definition by myself?

Nope! It's done like that to avoid the need for an actual base virtual class. In this way, all methods are perfectly known at build time, and can even be inlined (typically).

Please, just add the new method to the specializations (I may have missed some, please do a search):

mrpt/libs/maps/include/mrpt/maps/CPointsMap.h

Line 1178 in 46d80fc

class PointCloudAdapter<mrpt::maps::CPointsMap>

mrpt/libs/opengl/include/mrpt/opengl/pointcloud_adapters.h

Line 49 in deca624

class PointCloudAdapterHelperNoRGB

mrpt/libs/maps/include/mrpt/maps/CColouredPointsMap.h

Line 361 in 46d80fc

class PointCloudAdapter<mrpt::maps::CColouredPointsMap>

mrpt/libs/maps/include/mrpt/maps/CWeightedPointsMap.h

Line 162 in 46d80fc

class PointCloudAdapter<mrpt::maps::CWeightedPointsMap>

Update: my fault. I didn't realize you added the methods to PointCloudAdapter<mrpt::maps::CPointsMap>, it seemed to me that the method was added to CPointsMap directly! So, it's good!

jlblancoc · 2018-06-24T06:45:25Z

libs/maps/include/mrpt/maps/CSimplePointsMap.h

@@ -177,6 +177,10 @@ class PointCloudAdapter<mrpt::maps::CSimplePointsMap>
 	inline size_t size() const { return m_obj.size(); }
 	/** Set number of points (to uninitialized values) */
 	inline void resize(const size_t N) { m_obj.resize(N); }
+	/** Does nothing as of now */
+	inline void setDimensions(const size_t& height, const size_t& width)


same here...

jlblancoc · 2018-06-24T06:46:15Z

libs/obs/include/mrpt/obs/CObservation3DRangeScan.h

 	T3DPointsProjectionParams()
 		: takeIntoAccountSensorPoseOnRobot(false),
 		  robotPoseInTheWorld(nullptr),
 		  PROJ3D_USE_LUT(true),
 		  USE_SSE2(true),
-		  MAKE_DENSE(true)
+		  MAKE_DENSE(true),
+		  MAKE_ORGANIZED(false)


Forgot to add this,
MAKE_DENSE does not actually preserve the "organization" of the point cloud. It is only used to control whether or not the points can have nan / undefined values. If MAKE_DENSE is true, the cloud simply cannot contain undefined points and all the points contain finite values.

jlblancoc · 2018-06-24T06:46:21Z

libs/obs/include/mrpt/obs/CObservation3DRangeScan.h

@@ -828,6 +832,11 @@ class PointCloudAdapter<mrpt::obs::CObservation3DRangeScan>
 		if (N) m_obj.hasPoints3D = true;
 		m_obj.resizePoints3DVectors(N);
 	}
+	/** Does nothing as of now */
+	inline void setDimensions(const size_t& height, const size_t& width)


same here...

jlblancoc · 2018-06-24T06:46:36Z

libs/opengl/include/mrpt/opengl/CPointCloud.h

@@ -334,6 +334,10 @@ class PointCloudAdapter<mrpt::opengl::CPointCloud>
 	inline size_t size() const { return m_obj.size(); }
 	/** Set number of points (to uninitialized values) */
 	inline void resize(const size_t N) { m_obj.resize(N); }
+	/** Does nothing as of now */
+	inline void setDimensions(const size_t& height, const size_t& width)


same here...

jlblancoc · 2018-06-24T06:46:43Z

libs/opengl/include/mrpt/opengl/CPointCloudColoured.h

@@ -299,6 +299,10 @@ class PointCloudAdapter<mrpt::opengl::CPointCloudColoured>
 	inline size_t size() const { return m_obj.size(); }
 	/** Set number of points (to uninitialized values) */
 	inline void resize(const size_t N) { m_obj.resize(N); }
+	/** Does nothing as of now */
+	inline void setDimensions(const size_t& height, const size_t& width)


same here...

ah! Please, add this change to the docs/doxygen-pages/changelog.h list of changes...

Made the changes you requested!
And did a git commit --amend, to make it one commit. Hope that's okay.

jlblancoc · 2018-06-26T11:58:37Z

libs/obs/include/mrpt/obs/CObservation3DRangeScan_project3D_impl.h

@@ -11,6 +11,9 @@

 #include <mrpt/core/round.h>  // round()

+#include <pcl/point_cloud.h>
+#include <pcl/point_types.h>
+


Ey! Red alarm here! :-)

We can't assume PCL is installed, since PCL is an optional dependency for MRPT.
I think you could remove those headers and all should build nice. Why did you add them?
For the templates for build, it's enough to have forward declarations, and in any case, the PCL headers included in the user code (the caller that includes this file).

So, please, remove those two includes.

Oops, sorry. Removed those two includes and checked the build. No issues.
Pushed.

jlblancoc

Once the PCL includes are removed, this can be merged.

* Added MAKE_ORGANIZED parameter to T3DPointsProjectionParams to be able to generate an organized pcl point cloud from a 2D range image * Appropriately set pcl cloud dimensions

jlblancoc · 2018-06-26T14:11:13Z

Good. Let's wait for the travis result...

jlblancoc · 2018-06-29T06:28:52Z

@jolting : Please, look at this...

Error response from daemon: Get https://registry-1.docker.io/v2/: unauthorized: incorrect username or password

I think we didn't change anything in the docker area (?).

On the AppVeyor build, it seems it timed out (sigh)...

jolting · 2018-07-02T17:16:29Z

@jlblancoc Travis isn't passing in env variables to PR. That's a setting on travis. I'm not so worried about leaking docker hub credentials in this way because that account is only used for builds. It's just a risk if we enable builds for PR from outside of the main repo.

karnikram mentioned this pull request Jun 23, 2018

GUI App for Automatic Extrinsic Calibration of Range and Visual Sensors (Karnik Ram R) MRPT/GSoC2018-discussions#2

Open

jlblancoc requested changes Jun 24, 2018

View reviewed changes

karnikram force-pushed the master branch 2 times, most recently from b5267af to e8df091 Compare June 26, 2018 11:02

jlblancoc reviewed Jun 26, 2018

View reviewed changes

jlblancoc requested changes Jun 26, 2018

View reviewed changes

karnikram force-pushed the master branch from e8df091 to 012c399 Compare June 26, 2018 12:19

Add MAKE_ORGANIZED parameter

82110e8

* Added MAKE_ORGANIZED parameter to T3DPointsProjectionParams to be able to generate an organized pcl point cloud from a 2D range image * Appropriately set pcl cloud dimensions

karnikram force-pushed the master branch from 012c399 to 82110e8 Compare June 26, 2018 12:24

jlblancoc approved these changes Jun 29, 2018

View reviewed changes

jlblancoc merged commit 1f8f3ac into MRPT:master Jul 2, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add MAKE_ORGANIZED parameter #789

Add MAKE_ORGANIZED parameter #789

karnikram commented Jun 23, 2018 •

edited

jlblancoc left a comment

jlblancoc Jun 24, 2018

karnikram Jun 24, 2018

jlblancoc Jun 24, 2018

karnikram Jun 24, 2018

karnikram Jun 24, 2018

jlblancoc Jun 24, 2018

jlblancoc Jun 26, 2018

jlblancoc Jun 24, 2018

jlblancoc Jun 24, 2018

karnikram Jun 24, 2018

jlblancoc Jun 24, 2018

jlblancoc Jun 24, 2018

jlblancoc Jun 24, 2018

jlblancoc Jun 24, 2018

karnikram Jun 24, 2018

jlblancoc Jun 26, 2018

karnikram Jun 26, 2018 •

edited

jlblancoc left a comment

jlblancoc commented Jun 26, 2018

jlblancoc commented Jun 29, 2018

jolting commented Jul 2, 2018

Add MAKE_ORGANIZED parameter #789

Add MAKE_ORGANIZED parameter #789

Conversation

karnikram commented Jun 23, 2018 • edited

PR Description

jlblancoc left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

karnikram Jun 26, 2018 • edited

Choose a reason for hiding this comment

jlblancoc left a comment

Choose a reason for hiding this comment

jlblancoc commented Jun 26, 2018

jlblancoc commented Jun 29, 2018

jolting commented Jul 2, 2018

karnikram commented Jun 23, 2018 •

edited

karnikram Jun 26, 2018 •

edited