HERON Researchers Achieve 2nd Place in Two International Challenges at CVPR 2026 and MICCAI 2026

We are thrilled to announce that HERON researchers achieved 2nd place in two prestigious international challenges, held in conjunction with CVPR 2026 (IEEE/CVF Conference on Computer Vision and Pattern Recognition) and MICCAI 2026 (International Conference on Medical Image Computing and Computer-Assisted Intervention).

The first achievement came at the NTIRE Video Saliency Prediction Challenge, held on June 3, 2026, during the NTIRE Workshop at CVPR 2026. The team, consisting of N. Chaldaiopoulos, N. Efthymiou, A. Zlatintsi, P. Filntisis, K. Pastra, and P. Maragos, ranked 2nd with their proposed video saliency prediction system, which also established a new state-of-the-art on the DHF1K benchmark. The approach is described in the paper “PredJSal: Video Saliency via Predictive Self-Supervised Representations.”

The publication can be found by clicking here.

The second achievement came at the TIGER SQ-AI Challenge 2026, held on September 27, 2026, during the Endoscopic Vision Challenge at MICCAI 2026. The team, consisting of K. Spathis, G. Chatzichristodoulou, N. Kardaris, and K. Papadimitriou, achieved 2nd place in semantic segmentation (31 Classes) with their proposed deep learning system for surgical scene segmentation. The system demonstrated strong and consistent performance, ranking second in 88% of the bootstrap evaluation samples. The approach is described in the work “A Two-Family Ensemble of Surgical-Video CNN/ViT Segmenters and a Hierarchical DINOv2 ViT-L for Thoracic Anatomy Segmentation.”

More details about the proposed approach can be found here.

These achievements reflect the high quality of HERON’s research and its contributions to advancing artificial intelligence, computer vision, and medical image analysis, demonstrating the international recognition of its research efforts at prestigious and highly competitive scientific events.