SAGE: Shape-Adaptive Gated Experts for Medical Image Segmentation
A mixture-of-experts framework for medical histopathology segmentation with shape-adaptive routing and efficient expert specialization.
Selected work
A collection of research, engineering, and applied AI projects. Each card highlights the question, tools, and outcome behind the work.
A mixture-of-experts framework for medical histopathology segmentation with shape-adaptive routing and efficient expert specialization.
A systematic study of equivariance-preserving downsampling for low-resolution CNN classification, reaching strong CIFAR-10 accuracy with a compact model.
Exploring robust multi-modal representations for 3D perception by combining complementary LiDAR and camera signals.
A study of controllable image generation for emoji-style visual content using Stable Diffusion.
An implementation study of conditional flow matching for text-guided image synthesis.
A lightweight image classification application supporting multiple datasets and interactive inference.
An interpretable image segmentation pipeline based on color-space clustering with K-Means.
A document question answering prototype that combines PDF parsing, vector search, and LangChain.