- [08/2021] Article: Knowledge Evolution in Neural Networks
- [07/2021] Article: Cross-Iteration Batch Normalization
- [07/2021] Milestone: I am a research scientist at WhiteRabbit.AI
- [05/2021] Milestone: My journey at the University of Maryland has reached its end; I received a Ph.D. degree in Computer Vision. I have been lucky to have both Prof Larry Davis and Abhinav Shrivastava as my Ph.D. supervisors.
- [05/2021] Article: L2-CAF: A Neural Network Debugger
- [04/2021] Article: Deep Metric Learning Beyond Binary Supervision
- [03/2021] Defended my PhD dissertation 💪 🎉🎉
- [03/2021] Paper: One Oral Paper accepted in CVPR 2021
- [01/2021] Github: PyTorch implementation of Exploring Simple Siamese Representation Learning
- [12/2020] Article: Mining on Manifolds: Metric Learning without Labels
- [12/2020] Github: Tensorflow implementation of Representation Learning by Learning to Count
- [09/2020] Article: A Generic Visualization Approach for Convolutional Neural Networks
- [07/2020] Paper: One Paper accepted in ECCV 2020
- Full News list
||Feature Embedding, Metric Learning, Deep Networks, Machine Learning, Image segmentation, Texture classification, Patch matching.
||Python, C/C++, JAVA, OpenCV, MATLAB, mex files, and CUDA
TensorFlow, PyTorch, Keras, OpenCV, SimpleITK, CAFFE
Computer Science PhD (+MS) GPA: 4.0/4.0 - University of Maryland
Masters of Business Administration (Marketing Major) GPA 3.83/4.0 - Arab Academy for Science and Technology
Computer Science BS GPA 3.81/4.0 - Alexandria University- Faculty of Engineering
(click on image to expand)
|SVMax: A Feature Embedding Regularizer, arXiv 2021Ahmed Taha, Alex Hanson, Abhinav Shrivastava, Larry Davis
|Knowledge Evolution in Neural Networks, CVPR Oral 2021 (acceptance rate 24% -- Oral 4%)Ahmed Taha, Abhinav Shrivastava, Larry Davis
|A Generic Visualization Approach for Convolutional Neural Networks, ECCV 2020 (acceptance rate 27%)Ahmed Taha, Xitong Yang, Abhinav Shrivastava, Larry Davis
|Boosting Standard Classification Architectures Through a Ranking Regularizer, WACV 2020Ahmed Taha, Yi-Ting Chen, Teruhisa Misu, Abhinav Shrivastava, Larry Davis
|Unsupervised Data Uncertainty Learning in Visual Retrieval Systems, arXiv 2019Ahmed Taha, Yi-Ting Chen, Teruhisa Misu, Abhinav Shrivastava, Larry Davis
|Exploring Uncertainty in Conditional Multi-Modal Retrieval Systems, arXiv 2019Ahmed Taha, Yi-Ting Chen, Teruhisa Misu, Larry Davis
||Segmentation of Renal Structures for Image-Guided Surgery, MICCAI 2018 Junning Li, Pechin Lo, Ahmed Taha, Hang Wu, Tao Zhao
|Kid-Net: Convolution Networks for Kidney Vessels Segmentation from CT-Volumes, MICCAI 2018Ahmed Taha, Pechin Lo, Junning Li, Tao Zhao
|Two Stream Self-Supervised Learning for Action Recognition CVPRW 2018 Ahmed Taha, Moustafa Meshry, Xitong Yang, Yi-Ting Chen, Larry Davis
(DeepVision)- Extended Abstract Github Code
|Texture Synthesis with Recurrent Variational Auto-Encoder, ARXIV 2017Rohan Chandra, Sachin Grover, Kyungjun Lee, Moustafa Meshry, Ahmed TahaGithub Code
|Seeded Laplacian: An Interactive Image Segmentation Approach using Eigenfunctions, ICIP 2015Ahmed Taha, Marwan TorkiGithub Code
|Multi-Modality Feature Transform: An Interactive Image Segmentation Approach, BMVC 2015Moustafa Meshry, Ahmed Taha, Marwan TorkiCode
- Summer 2019: Student Associate at Honda Research Institute (HRI-US)
- Summer 2018: Research Assistant in University Of Maryland, sponsored by Honda Research Institute
- Summer 2017: Medical Image Analysis/Machine Learning Intern at Intuitive Surgical Inc
- Summer 2016: Emerging Graphics Group Intern at Adobe Systems Inc
[Summer 2018] Video Retrieval System [Honda Research Institute (HRI-US) internship]: Propose a descriptive markup language to describe participants' (e.g., cars and pedestrian) movements at road intersections. Using the proposed language, a deterministic polynomial-time algorithm is utilized to quantitatively compute an interpretable similarity metric between different driving intersection scenarios. This enables us to develop a video retrieval system for both stop-sign and traffic-light controlled intersections. We investigate multiple approaches to automatically transform trimmed autonomous navigation ego-videos at intersections into the proposed markup language.
- [Spring 2018-Summer2019] Autonomous Navigation Recognition & Retrieval: Sponsored by Honda Research Institute, I explore self-supervised approaches for ego-motion action recognition. I studied also video embedding, triplet loss retrieval and uncertainty estimation. This work led to a CVPRW2018 publication and pending anonymous submission.
- [Summer 2017] Semantic Segmentation [MICCAI Poster]: During Intuitive Surgical internship, I applied machine learning techniques to segment key anatomical structures from volumetric CT-images, in a fully automatic and semi-automatic fashion. We propose a convolution neural network developed using Keras, Tensorflow and Python libraries. Based on that work, two MICCAI 2018 papers are published.
- [Summer 2016] Patch Matching [Adobe Research Intern Expo Poster][Sample Video, Video 2]: During Adobe summer internship, I worked on a new selection/segmentation tool based on patch matching. The intuition is much similar to "Pseudo-polar based estimation of large translations rotations and scalings in images" but instead of comparing images, we compare patches. The selection results are remarkable, yet it suffered large computational time. Thus, it is not ready for industry use yet. While at Adobe, I intensively learned about Coherency Sensitive Hashing, The Generalized PatchMatch Correspondence Algorithm and of course Fourier transform. After the internship, I did additional evaluation against PatchMatchGraph: Building a Graph of Dense Patch Correspondences for Label Transfer. That's why I became expert with Darwin Framework.
- [Winter 2016] Texture Classification: I worked on a new texture classification approach classifying textures suffering a dark shadow. The results achieved was not good enough for a top conference. Yet, I reviewed the literature in great details. I implemented both local binary pattern (LBP) and various filter banks like Leung-Malik, Schmid, and maximum response. I got familiar with compressed sensing for texture classification. I ran my evaluation experiments using the following texture datasets: CUReT, UIUC and DTD.
-  Image Segmentation: Developed an approached for solving interactive image segmentation problem. The approach supports different user annotation forms like scribble, complete and incomplete trimaps, tight contour and bounding box. Qualitative and quantitative results are compared against Grabcut, Geodesic Star Convexity and MILCut.
- WACV Doctoral Consortium Plus Travel Award 2020.
- Graduate School's Outstanding Teaching Assistant Award for AY 2019-20 (Awarded to 2%).
- Gifted unrestricted 2500$ from Adobe Systems, Inc.
- University of Maryland Graduate School Deans Fellowship, 2015 and 2016
- Team has been chosen as one of the Young Innovators Awards(YIA) Program winners for the academic year 2008/2009.
- Awarded four successive times in college for the Excellent grade
(click to see course description)
- CMSC216 (Introduction to Computer Systems - Using C)
Machine representation of data including integers and floating point. Modern computer architectural features and their interaction with software (registers, caches). Interaction between user programs and the OS: system class, process, and thread management. Optimizing software to improve runtime performance using both compilers and hand turning.
- CMSC132 (Object-Oriented Programming II - Using JAVA)
Introduction to use of computers to solve problems using software engineering principles. Design, build, test, and debug medium -size software systems and learn to use relevant tools. Use object-oriented methods to create effective and efficient problem solutions. Use and implement application programming interfaces (APIs).
- CMSC420 (Data Structures - Using JAVA)
Description, properties, and storage allocation of data structures including lists and trees. Algorithms for manipulating structures. Applications from areas such as data processing, information retrieval, symbol manipulation, and operating systems.
- CMSC426 (Computer Vision - Using PYTHON)
An introduction to basic concepts and techniques in computer vision. This includes low-level operations such as image filtering and edge detection, 3D reconstruction of scenes using stereo and structure from motion, and object detection, recognition and classification.
PostscriptAfter earning my Bsc, I spent some time developing mobile apps for iOS. I co-founded Inova, a software development company.