Mengwei Ren

I am a Research Scientist at Adobe. I obtained my Ph.D. in Computer Science from NYU (2023), supervised by Prof. Guido Gerig under Visualization and Data Analytics Research (VIDA) Center . I have been fortunate to intern at Adobe, Google Research, and Siemens Healthineers.

My research lies at the intersection of computer vision and deep learning, with a focus on developing robust and controllable generative models for diverse applications in image editing, restoration, and composition.

Email  /  CV  /  Google Scholar  /  LinkedIn  /  Github

Full publication list is available here.

We explore portrait relighting by training a diffusion model to re-render synthetic faces.

We introduce a high-fidelity portrait shadow removal model that can effectively enhance the image of a portrait by predicting its appearance under disturbing shadows and highlights.

We introduce Relightful Harmonization, a lighting-aware diffusion model designed to seamlessly harmonize sophisticated lighting effect for the foreground portrait using any background image.

Image-conditioned Diffusion Probablistic Models (icDPMs) for restoration work well on benchmarks but not real images. We introduce a simple yet effective structure guidance that leads to significantly better visual quality on unseen images.

CNN-extracted global and local features are limited in capturing long-range spatial dependencies that are essential in biological anatomy. We present a keypoint-augmented fusion layer that extracts representations preserving both short- and long-range self-attention in a self-supervised manner.

We propose a local and multi-scale spatiotemporal representation learning method for image-to-image architectures trained on longitudinal (non i.i.d) images.

Based on an underlying assumption that morphological shape is consistent across imaging sites, we propose a segmentation-renormalized image translation framework to reduce inter-scanner heterogeneity while preserving anatomical layout.

We propose a generative adversarial translation framework for high-quality DW image estimation with arbitrary Q-space sampling given commonly acquired structural images.

We reformulate deformable registration and conditional template estimation as an adversarial game wherein we encourage realism in the moved templates with a generative adversarial registration framework conditioned on flexible image covariates.

cr: Jon Baron