Ensemble of Vision Encoders Wins Second Place in ICRA 2026 Segmentation Challenge

Researchers have developed a pretraining-diverse ensemble of foundation vision encoders for the ICRA 2026 GOOSE 2D Fine-Grained Semantic Segmentation Challenge. Their approach combines encoders like DINOv3, SigLIP2, and InternImage with a Mask2Former decoder, employing extensive training schedules and augmentation techniques. This ensemble achieved second place in the challenge, scoring 75.40% composite mIoU, and highlighted the pretraining recipe as the key factor for accuracy over model size or decoder design. AI

IMPACT Demonstrates effective ensemble techniques for robust outdoor scene understanding in computer vision.

RANK_REASON Technical report detailing a solution for a specific academic challenge. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• DINOv3
• Goose
• ICRA 2026 GOOSE 2D Fine-Grained Semantic Segmentation Challenge
• InternImage
• Mask2Former
• SigLIP2

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →