Research

I'm interested in computer vision, machine learning, optimization, image processing, and deep learning. Much of my research is about image restoration tasks such as single image super-resolution, image deblurring, and image denoising. Representative papers are highlighted here. Go to my Google Scholar profile for the full list.

Conference Publications

Imbalanced Domain Generalization for Robust Single Cell Classification in Hematological Cytomorphology
Rao Muhammad Umer, Armin Gruber, Sayedali Shetab Boushehri, Christian Matek, Carsten Marr
ICLR, 2023   (Spotlight presentation)

Paper Poster Presentation

We train a robust CNN for White Blood Cell (WBC) classification by addressing cross-domain data imbalance and domain shifts.

Raw image space improves single-cell classification in Acute Myeloid Leukemia
Rao Muhammad Umer, Christian Matek, Carsten Marr
Helmholtz Imaging Conference, 2022   (Appeared as a poster display)

Poster

We rethink the CNN classifier training in Raw image space instead of traditional RGB color space, diminish the illumination and coloring effects due to translate into the raw linear color space, and thus improve the classifiation performance.

SRResCSinGAN: Real Image Super-Resolution using GAN through modeling of LR and HR process
Rao Muhammad Umer, Christian Micheloni
AVSS, 2022   (Oral Presentation)

Paper Video

Training the GAN-based SRResCSinGAN with learnable adaptive sinusoidal nonlinearities incorporated in LR and SR models by directly learn degradation distributions and then synthesize paired LR/HR training data to train the generalized SR model to real image degradations.

RBSRICNN: Raw Burst Super-Resolution through Iterative Convolutional Neural Network
Rao Muhammad Umer, Christian Micheloni
NeurIPS, 2021   (Appeared in the Machine Learning and Physical Sciences workshop)

Paper Poster code Video

Star Fork

Training the RBSRICNN for Raw Burst Super-Resolution task that restore the clean RGB SR image by following the burst photography pipeline as a whole by a forward (physical) model.

SR2*GAN: A Deep Residual Star Generative Adversarial Network for multi-domain Image Super-Resolution
Rao Muhammad Umer, Asad Munir, Christian Micheloni
SPLITECH, 2021   (Oral Presentation)

Paper Video

Training the SR2*GAN in a StarGAN like network topology with a single model to super-resolves the LR images for the multiple LR domains.

BSRICNN: Deep Iterative Convolutional Neural Network for Raw Burst Super-Resolution
Rao Muhammad Umer, Christian Micheloni
CVPR, 2021   (Participated in the NTIRE 2021 Burst SR challenge)

Factsheet Burst-SR-Report

Training the BSRICNN for Burst Super-Resolution that solves the Burst SR task by iterative refinement of the intermediate SR estimates.

ISRResDNet: A Light-weight Deep Iterative Residual Convolutional Network for Super-Resolution
Rao Muhammad Umer, Christian Micheloni
ECCV, 2020   (Participated in the AIM 2020 Efficient SR challenge)

Factsheet Efficient-SR-Report

Training the ISRResDNet for Efficient Super-Resolution that solves the SR task as a sub-solver of the image denoising by the residual denoiser networks.

SRResCycGAN: Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real-Image Super-Resolution
Rao Muhammad Umer, Christian Micheloni
ECCV, 2020   (Participated in the AIM 2020 Real-Image SR challenge)

Paper RISR-Report code Video Project

Star Fork

Training the SRResCycGAN network for Real-Image Super-Resolution with a generative adversarial network (GAN) framework for the LR to HR domain translation in an end-to-end manner.

ISRResCNet: Deep Iterative Residual Convolutional Network for Single Image Super-Resolution
Rao Muhammad Umer, Gian Luca Foresti, Christian Micheloni
ICPR, 2020   (Oral Presentation)

Paper code Video Project

Star Fork

Training the ISRResCNet network in an iterative manner with a residual learning approach for super-resolution to follow the image observation (physical) model.

SRResCGAN: Deep Generative Adversarial Residual Convolutional Networks for Real-World Super-Resolution
Rao Muhammad Umer, Gian Luca Foresti, Christian Micheloni
CVPR, 2020   (Participated in the Ntire2020 RWSR challenge)

Paper RWSR-Report code Video Project

Star Fork

Training the SRResCGAN network for super-resolution to follow the real-world degradation settings by adversarial training the model with pixel-wise supervision in the HR domain from its generated LR counterpart.

SRWDNet: Deep Super-Resolution Network for Single Image Super-Resolution with Realistic Degradations
Rao Muhammad Umer, Gian Luca Foresti, Christian Micheloni
ICDSC, 2019   (Oral Presentation, Best Runner-up Paper Award)

Paper Presentation

Training the SRWDNet for super-resolution that works for blur kernels of different sizes and different noise levels in an unified residual CNN-based denoiser network.

Journal Publications

Blockchain-Based Swarm Learning for the Mitigation of Gradient Leakage in Federated Learning
Hussain Ahmad Madni, Rao Muhammad Umer, Gian Luca Foresti
IEEE Access, 2023

Paper

In this work, we proposed the method for Blockchain-Based Swarm Learning for the mitigation of gradient leakage in federated learning.

Swarm-FHE: Fully Homomorphic Encryption based Swarm Learning for Malicious Clients
Hussain Ahmad Madni, Rao Muhammad Umer, Gian Luca Foresti
International Journal of Neural Systems, 2023

Paper

In this work, we proposed the method Swarm-FHE for Fully Homomorphic Encryption based Swarm Learning for Malicious Clients.

Theses and Dissertation

Deep Convolutional Neural Networks for Image Super-Resolution
Rao Muhammad Umer
Ph.D. dissertation, Department of Industrial and Information Engineering, University of Udine (UNIUD), Udine, Italy, 2021

Dissertation Presentation Project

In this work, we describe the algorithmic advances and results obtained by the proposed methods in the image Super-Resolution field.

Deep Web Extractor (DWX): Content Discovery From Deep Web Using Large Scale Data Analytics Paradigm
Rao Muhammad Umer
DCIS,PIEAS, 2016   (Best Thesis Award)

Thesis code Project

Deep Web Extractor system is a cloud-based web application that uses machine learning techniques for crawling and data discovery from the Deep Web (i.e., massive and quality portion of World Wide Web) to build knowledge based databases.