Overall feedback

Very pleasantly surprised by how you approached the assignment. Most of you did a great job, including some very creative solutions.

Highlights / common positive points

• using IfcConvert’s options (eg --weld-vertices or --orient-shells) to fix some errors, reduce postprocessing or help in debugging
• attempting to use original geometries when possible, falling back on other options when issues arise (eg invalid or very complex geometries)
• using a convex hull to fix geometries or to close space left by windows/doors
• using Minkowski sum to close small gaps or remove small objects
• finding a good BIM model (eg reasonably valid and with simple-ish geometry) that still met the homework requirements
• using the original geometry locations to assign semantics
• using hints provided by your classmates or found online (with attribution)

Common mistakes / limitations

• misunderstanding the OBJ files produced by IfcConvert, particularly how shells are defined through usemtl rather than s
• not using an incremental approach (start from simple model and move on from there) -> resulted in complex code that isn’t really working
• searching for things in vectors/sets/maps rather than storing the (known) location of elements
• re-implementing existing CGAL functionality, like computation of planes and normals
• not validating output (cjval / val3dity) -> helps to detect errors and to create better output
• shallow analysis of the results -> not showing limitations of method

How I marked Homework 3

First, I read your report to assign rough marks for each of the categories. If I needed more information, I checked your submitted datasets and code for reference. Afterwards, I opened your CityJSON file in azul to check whether the required objects are present (with their geometry and semantics). Finally, I adjusted the marks according to the schema below.

followed all rules [1]

• 1 - all good

quality of results [6]

• 6 - complete pipeline that works on a model of significant complexity (eg Duplex or KIT models) with very good results
• 5 - complete pipeline that works only with simpler model (eg IfcOpenHouse)
• 4 - pipeline mostly working but with some issues (eg no separation of interior/exterior or no rooms)
• 3 - pipeline half working

quality of report [3]

• 3 - very good report
• 2 - good report but with some issues (eg shallow assessment of results)
• 1 - significant parts missing in report (eg description of key steps or results)

delays [negative]

• 0 - no delay or delays of a few minutes (no penalty)
• -1 - delays of up to a day (with tolerance of a few minutes too)