Skip to content

Releases: IGNF/lidar-prod

V1.9.9 - PATCH - Bump PDAL to v2.5.1 fixing overlay issue

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 15 Feb 07:01

TL;DR: Building Completion includes all points near already confirmed building, not only candidates. Building module logic is relativelly stable by now. Some PDAL bug is corrected.

What's Changed since V1.8

New Contributors

Full Changelog: V1.6.2...V1.9.9

M11.1 Building completion puts non-candidate in new group

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 04 Nov 12:42
da40eaf
  • Completion includes all points near already confirmed building, not only candidates. The non-candidates will form a new group (ID=1) and groups identified in BuildingIdentification task will be shifted to start at ID=2.
  • Identification will ignore already confirmed points, non-candidates, and the points in the group of ID=1.

As a consequence, Building identification somehow starts within task Building Completion, as the first group of suggested building is found withing Building Completion.

The doc illustrates this is in a nice png.

In addition:

  • relaxation factor of threshold is removed for these tasks, as it was not used.
  • Dimensions Normal, Deviation, Distance are protected from deletion, and saved as Unsigned Integers for now.

What's Changed

Full Changelog: M10.0...M11.1

M10.0 Reorder use of entropy information

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 03 Nov 15:32
05f58db

Entropy is now activly used, since BD Uni allows for validation even though the entropy of AI may be high.

Until now the optimized thresholds for high entropy was very high due to a order that forbid BD Uni to catch these high entropy group.

What's Changed

Full Changelog: M9.0...M10.0

M9.0 Refactor + Vegetation/Unclassifed Module

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 29 Sep 09:23
a969533

Large refactor ayant une influence sur l'API.
Ajout d'un module générique de seuillage optimisé pour la création d'un groupe (dimension "Group" du format LAS lisible sous TerraScan), utilisé pour les classes Végétation et Non-Classé.
Push automatique d'image Docker vers Nexus pour versionnement.
Fix fonctionnel sur la complétion de bâti.

What's Changed

  • Module d'identification de la végétation et des non-classés by @MichelDaab in #62
  • PATCH: fix mermaid diagramm display by @MichelDaab in #64
  • PATCH: Fix unclustered_candidates_mask for building completion by @CharlesGaydon in #66
  • Push docker images to a Nexus Repository within CICD workflow by @MichelDaab in #67
  • PATCH - Remove unnecessary assertion that prevented optimization tasks. by @CharlesGaydon in #68

Full Changelog: M8.4...M9.0

V1.6.2 - laspy with a LAZ backend

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 30 Aug 14:18
00d8b32

Simple change of environment to support reading/writing LAZ with laspy.

What's Changed

Full Changelog: V1.6.0...V1.6.2

V1.6.0 - Tests and Documentation

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 05 May 16:19

Important refactor coupled with more continuous integration/deployment and a good high-to-medium level test coverage.
Docker deployment as new tagged images for pushes to dev, main and prod branches.
Documentaton depoyed to github pages.

What's Changed

Full Changelog: V1.3.0...V1.6.0

M8.4

M8.4 Pre-release
Pre-release

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 28 Mar 15:33
e102738

Implements the use of entropy to flag unsure candidate building groups.
This is slightly different than relying on single probability when we predict more than 2 classes. Furthermore, it allows to flag as unsure elements that have both a portion of them with high certainty (positive or negative) and also a portion of them with high entropy.

M8.3

M8.3 Pre-release
Pre-release

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 24 Mar 10:13
27d048d

This enable to optimize a relaxation factor that is applied to minimal confidence thresholds under BDUni.
Automation bumps from 90.0% to 91.4%. But most importantly, this relaxation threshold is also used during building completion and during building identification, which effectively account for BDUni in these tasks.

What's Changed

Full Changelog: V1.1.0...M8.3

V1.1.0

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 21 Mar 10:11
e26c333

This fixes a channel name error that would make Building Identification identify isolated building candidates points (in e.g. walls) instead of ignoring them.
Output channels are better defined (Group, building). CI/CD is further improved using Docker.

What's Changed

Full Changelog: V1.0.0...V1.1.0

V1.0.0

Choose a tag to compare

@CharlesGaydon CharlesGaydon released this 28 Feb 16:11

Includes Building Validation + Building Identification + Dims Cleaning

What's Changed

Full Changelog: https://github.com/IGNF/lidar-prod-quality-control/commits/V1.0.0