S. Mahdi H. Miangoleh
I am a Computing Science PhD student at Simon Fraser University (SFU). I do research at Computational Photography Lab under the supervision of Prof. Yağız Aksoy.
I got my Master's degree in Computing Science at SFU in Summer 2022. My MSc Thesis is on Boosting Monocular Depth Estimation to High Resolution.
Before joining SFU I got my bachelor's degree in Electrical Engineering-Digital Systems at Sharif University of Technology.
News
- May, 2023 I started an Internship at Bosch AI under Yuliang Guo!
- Aug, 2022 I graduated with a Master's degree in Computing Science from Simon Fraser University (Thesis)! I will be continuing my studies toward a PhD degree under supervision of Prof. Yağız Aksoy.
- Aug, 2021 I started an Internship at Adobe Research working with Zoya Bylinskii!
Research
My research focuses on computational photography, 3D computer vision, learning-based approaches for vision and graphics, and realistic image editing. I also have experience in embedded systems and desktop front-end application development.
Depth Any Camera: Zero-Shot Metric Depth Estimation from Any Camera
Yuliang Guo, Sparsh Garg, S. Mahdi H. Miangoleh, Xinyu Huang, Liu Ren
Proc CVPR 2025
We introduce Depth Any Camera (DAC), a framework enabling superior zero-shot generalization in metric depth estimation for large FoV cameras, including fisheye and 360°. Tired of collecting new data for specific cameras? DAC maximizes the utility of every existing 3D data for training, regardless of the specific camera types used in new applications.
Scale-Invariant Monocular Depth Estimation via SSI Depth
S. Mahdi H. Miangoleh, Mahesh Reddy, Yağız Aksoy
Proc SIGGRAPH 2024
We present a novel approach that leverages SSI depth inputs to enhance SI depth estimation, streamlining the network's role and facilitating in-the-wild generalization for SI depth estimation while only using a synthetic dataset for training.
Intrinsic Harmonization for Illumination-Aware Compositing
Chris Careaga, S. Mahdi H. Miangoleh, Yağız Aksoy
Proc SIGGRAPH Asia 2023
We introduce a self-supervised illumination harmonization approach formulated in the intrinsic image domain. First, we estimate a simple global lighting model from mid-level vision representations to generate a rough shading for the foreground region. A network then refines this inferred shading to generate a harmonious re-shading that aligns with the background scene.
Realistic Saliency Guided Image Enhancement
S. Mahdi H. Miangoleh, Zoya Bylinskii, Eric Kee, Eli Shechtman, Yağız Aksoy
Proc CVPR 2023
We train and expliot a problem specific realism network to train a saliency-guided image enhancement network which allows maintaining high realism across varying image types while attenuating distractors and amplifying objects of interest. Our proposed approach offers a viable solution for automating image enhancement and photo cleanup operations.
Boosting Monocular Depth Estimation Models to High-Resolution via Content-Adaptive Multi-Resolution Merging
S. Mahdi H. Miangoleh*, Sebastian Dille*, Long Mai, Sylvain Paris, Yağız Aksoy
Proc CVPR 2021
We demonstrate that there is a trade-off between a consistent scene structure and the high-frequency details, and merge low- and high-resolution estimations to take advantage of this duality using a simple depth merging network and generate multi-megapixel depth maps with a high level of detail using a pre-trained model.
Interactive Editing of Monocular Depth
Obumneme Stanley Dukor, S. Mahdi H. Miangoleh, Mahesh Kumar Krishna Reddy, Long Mai, Yağız Aksoy
Proc SIGGRAPH POSTER 2022
In this work, we present a lightweight, web-based interactive depth editing and visualization tool that adapts low-level conventional image editing operations for geometric manipulation to enable artistic control in the 3D photography workflow.
Theses
Boosting Monocular Depth Estimation to High Resolution
MSc Thesis
Seyed Mahdi Hosseini Miangoleh
We demonstrate that there is a trade-off between a consistent scene structure and the high-frequency details, and merge low- and high-resolution estimations to take advantage of this duality using a simple depth merging network and generate multi-megapixel depth maps with a high level of detail using a pre-trained model.