GOOSE 2D Fine-Grained Semantic Segmentation Challenge
1st Place at ICRA 2026 Workshop on Field Robotics
The DGIST–MIT joint research team placed 1st among 56 international teams in the GOOSE 2D Fine-Grained Semantic Segmentation Challenge at ICRA 2026.
The team was led by Professor Sung-Hoon Yoon and included Jaeil Park, Hyobin Choi, and Sangjin Lee, with co-advising from Dr. Hyungtae Lim of the MIT SPARK Lab.
Key Achievement
The challenge evaluated semantic segmentation across 64 fine-grained classes in complex, unstructured outdoor environments. The team combined DINOv3 with Mask2Former to improve robust scene understanding, particularly for rare objects with limited training data.
The technology can support autonomous vehicles, disaster-response robots, smart agriculture, and construction-site robots operating in challenging outdoor environments.