Monday, April 29, 2019

Create a Facebook 3D Photo of a digital elevation model using QGIS

Facebook has a 3D photo feature which brings perspective and movement to the Facebook news feed and timeline images. A live example can be seen embedded here:

Normally these images are created using RGB-D cameras, where D stands for depth. However it is also possible to workaround it without a RGB-D camera by uploading to Facebook two images - an RGB image and its corresponding grayscale depth image, and making sure the depth image has the same name as the RGB image with a '_depth' suffix. The depth image is grayscale only with values ranging from dark to light -  the dark pixels representing the furthest pixels and the light pixels representing the nearest pixels.

Using the ideas above, the following steps illustrate how to create a Facebook 3D Photo image from a SRTM digital elevation model (DEM) using QGIS.

  1. Startup QGIS. Create a new project. If the canvas background is not already black, select Project | Project Properties.

    The Project Properties dialog box appears.
  2. In the Background color field, change the color to black. Click OK.

    The canvas background becomes black.
  3. Drag and drop a DEM file onto the canvas e.g. srtm_54_07.tif.

  4.  Optional. To serve the purpose of having a background, an OpenStreetMap (OSM) layer is displayed behind the DEM layer.

  5. Optional. To display only DEM heights above an elevation of interest, e.g. 200 meters, right click on the DEM layer and choose Properties. Then add in a new transparent entry to the transparent pixel list from 0 to 200 meters, as shown below. Then click OK.

    The Layer Properties dialog box appears.

    The DEM display is updated.

  6. Make a duplicate of the DEM layer by right clicking on the DEM layer and choosing Duplicate. Rename the copy as 'depth' as it will serve to generate a depth image later on.

  7. Right click on the top most DEM layer and choose Properties. In the Layer properties dialog box, click Style.

  8. Change the Render Type to Hillshade. Click OK.

    The DEM layer is now rendered as a hill shade.
  9. Make a duplicate copy of the hill shaded DEM layer by right clicking the layer and choosing Duplicate. Rename the copy as 'hillshade' as shown below.

  10. Right click on the top most DEM layer and choose Properties. Select the Style tab.

  11. In the Rendering type field, choose Singleband pseudocolor. Choose a suitable color gradient e.g. Spectral and toggle Invert if necessary. Choose a suitable Blending mode e.g. Screen.

  12. Click OK.

    The DEM layer is rendered with a colored gradient.
  13. With the gradient colored, hill shaded and OSM layers displayed on and the depth layer displayed off, export out the canvas by selecting Project | Save as image and specifying an output file name e.g. srtm.png.

    The srtm image is generated.
  14. Now turn off the gradient, hill shaded and OSM layers. Turn on the depth layer as shown below.

    Select Project | Save as image to export out the depth image with a '_depth' suffix e.g. srtm_depth.png.

    The srtm_depth image is generated.
  15. Now open up Facebook in a browser and click the Post Photo/Video button.

    The File Upload dialog box appears.
  16. Browse and select the generated srtm.png and srtm_depth.png files. Click Open.

    Facebook generates the 3D Photo.

    A live example is available here at

Monday, April 22, 2019

Migrating Android support library Renderscript code to AndroidX

Android is deprecating the Support compatibility libraries  and replacing them with the new AndroidX libraries. To aid in the migration of Android projects using the deprecated support libraries, Android Studio has a Refactor to AndroidX tool. I used that to migrate an old Android project and while it works for the most part, it did not manage to migrate the files using the support renderscript classes; I presume that is because the AndroidX namespace does not have the equivalent renderscript classes.

To complete the migration to AndroidX, I had to do the following by hand:

  1. In all the code files using the* classes, replace them to android.renderscript.* as shown below.

  2. In the app's build.gradle file, upgrade the renderscriptTargetApi level to 17 and turn off the renderscript support mode (renderscriptSupportModeEnabled to false) as shown below.

  3. Save all the files and rebuild the project.

    The project should manage to sync and build successfully.
Note: At the time of writing, Android can only compile android.renderscript.* classes into 32 bit *.bc files. This may be a problem in future as Google Play is moving to apks that support 64 bit files. 

Monday, April 15, 2019

Speed up Android Studio startup by disabling unnecessary plugins

Android Studio loads a relatively large number of plugins on start up. Depending on your workflow, some of them may not be needed. For instance, if you don't use Subversion as your source code repository, then the Subversion plugin may not need to be loaded especially if your machine is slow or RAM is limited.

To disable plugins, do the following:

  1. Start Android Studio.

  2. In the Configure drop down menu at the bottom right, choose Plugins.

    The Plugins dialog box appears.

  3. Scroll down the list and toggle off the unneeded plugins e.g. CVS Integration. Click OK.

    The Prompt appears.

  4. Click Restart.

    Android Studio is restarted without the disabled plugins.

Monday, April 8, 2019

Segment individual trees from TLS point clouds with 3DForest

Identifying individual trees from point clouds scanned using Terrestrial LiDAR System (TLS can be easily segmented using this freely downloable software 3DForest. For more information about 3DForest, visit

Note: although 3DForest has its own ground classification filter, I found it better to perform the ground / non-ground classification using alternative software like PDAL.

To use 3DForest to segment individual trees, the following steps can be done:
  1. Start up 3DForest. Select Project | New Project to create a new project.

    The 3DForest application appears.
  2. In the menu, choose Project | Import | Import Terrain Cloud to load in a LAS file of only ground points.

  3. In the dialog box that pops up, choose a ground class only LAS file, e.g. sample_ground.las.

    The LAS file is converted into a PCD format file and displayed.
  4. Next, choose Project | Import Vegetation Cloud. Choose a non-ground class only LAS file, e.g. sample_vegetation.las.

    The non-ground class LAS file is converted into PCD format and displayed.
  5. In the menu, select Vegetation | Automatic Segmentation to run the individual tree segmentation algorithm.

    The Automatic Segmentation dialog box appears.

  6. In the Input Vegetation Cloud combo box, choose the newly loaded sample_vegetation LAS file.
  7. In the Input Terrain Cloud combo box, choose the loaded sample_ground LAS file.
  8. If necessary, change the parameters, e.g. Prefix of Clouds value from ID to TreeID. Click OK.

    The individual trees are segmented into individual point clouds.

  9. If required, select Project | Export | Export Clouds to export the segmented trees to text or PLY formatted files.