YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/model-cards#model-card-metadata)

S23DR 2025 Challenge: 4th Place Solution

This repository contains the 4th place solution for the S23DR 2025 Challenge. For more details, refer to the writeup included in this repository.

Approach

The solution is a fully handcrafted pipeline based on classical computer vision and geometry. It builds on the official baseline with improvements in edge detection, depth assignment, and 3D merging. No learned models were used.

Final Score

  • HSS Mean: 0.3309
  • Corner F1: 0.4310
  • Edge IOU: 0.2773

How to Run

The competition platform executes script.py to generate predictions. To run locally, you would need a dataset entry and could call the predict_wireframe function from solution.py.

File Structure

  • solution.py: Contains the main logic for the wireframe prediction pipeline.
  • script.py: The entry point script for the Hugging Face competition platform.
  • requirements.txt: Lists all Python dependencies.
  • hoho2025/: A helper module containing color mappings.
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support