New benchmark BOCCHI and MSDCT-UNet improve motion blur detection

Researchers have introduced BOCCHI, a new benchmark designed to improve local motion blur detection in computer vision. Unlike previous benchmarks, BOCCHI uses real-world imagery where sharp regions overlap with blur gradients, preventing models from exploiting shortcuts. The team also developed MSDCT-UNet, a frequency-aware model that incorporates multi-scale DCT priors via DCT Attention and FiLM, achieving top performance on BOCCHI and demonstrating strong cross-dataset transfer capabilities. AI

IMPACT This research could lead to more robust motion blur detection in real-world scenarios, improving image and video processing applications.

RANK_REASON The cluster describes a new academic paper introducing a benchmark and a model for a specific computer vision task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• Bocchi
• computer science
• Computer vision and pattern recognition
• DCT Attention
• FiLM
• MSDCT-UNet
• U-Net

• paper
• other

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