The Division of Arts and Machine Creativity (AMC) at The Hong Kong University of Science and Technology (HKUST) is proud to announce that Prof. Harry Yang, Assistant Professor of AMC, and his research team have received the prestigious CVPR 2026 Compute Transparency Champion Award for their paper, “DenDiff: Density-Guided Diffusion for Quantity-Aware Image Synthesis.”
The award-winning paper introduces DenDiff, a new approach for improving quantity control in text-to-image generation. While diffusion models can generate highly realistic images, they often struggle to produce the exact number of objects specified in text prompts. DenDiff addresses this challenge by introducing density-guided generation, using fine-grained visual features to create density maps that guide the image synthesis process toward the desired object count.
Experiments on the FSC-147 and ShanghaiTech benchmarks show that DenDiff achieves state-of-the-art counting accuracy while maintaining strong visual quality. By combining semantic image generation with explicit numerical and spatial control, the method enables more reliable and controllable image synthesis for real-world applications.
About CVPR
CVPR (Conference on Computer Vision and Pattern Recognition), jointly organized by the IEEE and the Computer Vision Foundation (CVF), is widely regarded as one of the world's premier conferences in computer vision and artificial intelligence. The conference serves as a leading venue for the presentation of groundbreaking research from academia and industry.
