Knowledge Base
Recent Changes

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision 		Previous revision
recordings:tutorials:videos:pointcloud_labeling [2024/02/06 15:03]
arnaud created 		recordings:tutorials:videos:pointcloud_labeling [2024/04/10 08:37] (current)
arnaud
Line 1: Line 1:
-====== Pointcloud labeling for training ======+====== AI-Labeling: Annotate pointcloud for AI training ======
  
  
-This video is about labeling pointcloud to serve custom AI-training jobs and create detectors to be used in the Analysis Engine AI. \\+This page is about labeling pointcloud to serve custom AI-training jobs and create detectors to be used in the Analysis Engine. \\
  
 ==== Step by step ==== ==== Step by step ====
  
-  - Open the pre-prepared Mobile Mapping Resource +  - Create a trajectory file that will be used as a canvas to easily navigate through the scene and label all elements of interest 
-  - Inspect the source dataset +      * Drag & drop your mapping run {{ :recordings:tutorials:videos:pointcloud_labeling_001.mp4?nolink&850x478 |}} 
-      * The image folder should not contain processed images +      * Create a vector resource for trajectory to constraint navigation {{ :recordings:tutorials:videos:pointcloud_labeling_002.mp4?nolink&850x478 |}} 
-      * Only .jpg image format is accepted  +      * Copy trajectory file in installation folder {{ :recordings:tutorials:videos:pointcloud_labeling_003.mp4?nolink&850x478 |}} 
-  - Download latest detectors from [[https://communities.bentley.com/products/3d_imaging_and_point_cloud_software/w/wiki/54656/context-insights-detectors-download-page|Context Insights detectors download page]] +      * Define trajectory as linear reference and lock it {{ :recordings:tutorials:videos:pointcloud_labeling_004.mp4?nolink&850x478 |}} 
-  - Launch the Analysis Engine +  - Create an asset class that will be labeled (e.g: pylons)  
-  - Start the  cracks detection  +      * Create asset ressource {{ :recordings:tutorials:videos:pointcloud_labeling_005.mp4?nolink&850x478 |}} 
-  - Import as objects the resulted detected image pixels/annotations +      * Pinpoint asset objects in mapping run {{ :recordings:tutorials:videos:pointcloud_labeling_006.mp4?nolink&850x478 |}} 
-  - Convert image annotations into objects +  - Select pointcloud to attribute to an asset object  
-  - Inspect the resulted objects+      * Select associated pointcloud and add pointcloud selection to asset objects {{ :recordings:tutorials:videos:pointcloud_labeling_007.mp4?nolink&850x478 |}} 
 +      * Measure next asset object until the entire dataset is reviewed {{ :recordings:tutorials:videos:pointcloud_labeling_008.mp4?nolink&850x478 |}} 
 +  - Generate consolidated classified pointcloud + assets to serve AI-detector training  
 +      * Generate data and merge pointcloud selections {{ :recordings:tutorials:videos:pointcloud_labeling_009.mp4?nolink&850x478 |}} 
 +      * Pointcloud selection to classification {{ :recordings:tutorials:videos:pointcloud_labeling_010.mp4?nolink&850x478 |}} 
 +  - Upload classified pointcloud to enable detector training  
 +      * Upload classified OPC to Reality Management {{ :recordings:tutorials:videos:pointcloud_labeling_011.mp4?nolink&850x478 |}} 
 +      * Bentley team takes it from here and will start detector training
  
- 
-==== Used Data ==== 
- 
-  * The Mapping Resource used for this video:  "ContextInsights cracks detection"  
-  * The instructions to download: [[recordings:tutorials:notes|]] 
- 
-==== Reference Documentation ==== 
- 
-  * [[237:desktop_ext:context_insights|]] 
-  * [[237:technology:platforms:install_contextinsights|]] 
-  * [[237:desktop_ext:image_annotation_object|]] 
 
Last modified:: 2024/02/06 15:03