Computer Laboratory

Aamir Mustafa

About Me

I am a PhD student at the Graphics & Interaction (Rainbow) Group, University of Cambridge, working under the supervision of Dr. Rafal K. Mantiuk. My current research is focused on Applications of Machine/Deep Learning in Computer Graphics. 

  LinkedIn    GitHub    Google Scholar    Academic CV

Running Projects - Visual Reports

 

Conditional training of Invertible Neural Nets: The report shows some initial results on the BBC documentary (Planet Earth-Jungles) for movie color grading using INNs.

Movie Tone Mapping using per frame meta-data: Ablation study on the different sizes of meta data required for SDR frame encoding.

Different movies: Best results on two different movies.

Movie Tone Mapping: Some initial results for end-to-end trained encoder-decoder models for image tone mapping.

Movie Tone Mapping - Histogram Matching: Results for training a network to learn histogram matching between HDR and SDR frames.

Study on Feature Encoding for Tone Mapping: Some results on the BBC documentary Planet Earth Episode 2 (Mountains).

Quick Updates

  • 1 paper accepted in London Imaging Meeting (LIM) 2022
  • 1 paper accepted in Winter Conference on Applications of Computer Vision (WACV) 2022
  • 1 paper accepted in European Conference on Computer Vision (ECCV) 2020
  • 1 paper accepted in IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI) 2020
  • 1 paper accepted in IEEE Transactions on Image Processing (TIP) 2020
  • 1 paper accepted in International Conference on Computer Vision (ICCV) 2019

Publications

A Comparative Study on the Loss Functions for Image Enhancement

Authors: Aamir Mustafa, Hongjie You, Rafal K. Mantiuk

London Imaging Meeting (LIM) 2022

  Paper

Training a Task Specific Image Reconstruction Loss

Authors: Aamir Mustafa, Aliaksei Mikhailiuk, Dan Andrei Iliescu, Varun Babbar, Rafal K. Mantiuk

Winter Conference on Applications of Computer Vision (WACV) 2022

Project Page   Paper  Code

Transformation Consistency Regularization - A Semi Supervised Paradigm for Image-to-Image Translation

Authors: Aamir Mustafa, Rafal K. Mantiuk

European Conference on Computer Vision (ECCV), 2020

Project Page   Paper  Supplementary Material  Code

Deeply Supervised Discriminative Learning for Adversarial Defense

Authors: Aamir Mustafa, Salman Khan, Munawar Hayat, Roland Goecke, Jianbing Shen, Ling Shao

IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 2020

Paper    Code

Adversarial Defense by Restricting the Hidden Space of Deep Neural Networks

Authors: Aamir Mustafa, Salman Khan, Munawar Hayat, Roland Goecke, Jianbing Shen, Ling Shao

International Conference on Computer Vision (ICCV), 2019

Paper   Code

Image Super-Resolution as a Defense against Adversarial Attacks

Authors: Aamir Mustafa, Salman Khan, Munawar Hayat, Jianbing Shen, Ling Shao

IEEE Transactions on Image Processing (TIP), 2020

Paper   Code

Prediction and Localization of Student Engagement in the Wild

Authors: Amanjot Kaur, Aamir Mustafa, Love Mehta, Abhinav Dhall

Digital Image Computing: Techniques and Applications (DICTA), 2018

Paper   Code

Heart Rate Estimation from Facial Videos for Depression Analysis.

Authors: Aamir Mustafa, Shalini Bhatia, Munawar Hayat, Roland Goecke

International Conference on Affective Computing and Intelligent Interaction (ACII), 2017 

Paper   Code

Experience

PhD Student - Computer Science Dept. University of Cambridge (Oct 2019 - Present)

Leveraging the use of semi-supervised learning for image to image translation.

Computer Vision Research Intern - Inception Institute of Artificial Intelligence (Sep 2018 - Sep 2019)

Working on making deep neural networks robust against adversarial attacks.

Computer Vision Research Intern - Indian Institute of Technology, Ropar (Dec 2017 - Mar 2018)

Worked on prediction and localization of student engagement in response to a stimuli video (e-learning environment) from facial expressions using Deep Multi-Instance Learning.

Machine Learning Research Intern - University of Canberra, Australia (Dec 2016 - Feb 2017)

Estimation of Heart rate of different individuals and its variations over the span of video from their facial videos by extracting plethysmograph (PG) signals from green channel of the frames. Considering heart rate as extracted feature, individuals are classified into two categories - healthy controls and depressed patients using a linear SVM classifier.

Completed Projects

 

Transformation Consistency Regularization

Perceptual Loss for Image Restoration

Contact

Email: am2806@cam.ac.uk , aamir.mustafa@cl.cam.ac.uk
Office Phone: +44 (0)1223 763561
Office: SS04, Computer Laboratory
Address: Computer Laboratory, William Gates Building, 15 JJ Thomson Ave, Cambridge CB3 0FD