Recruiter LogIn

Don't have an account? Register now

Forgot Password?

Recruiter Registration

Custom

Select Your Default Landing Page

Photo

Yaqub *********** Computer Vision Engineer

$720 / day

Summary

Experienced professional with a robust background in computer vision, machine learning, and medical imaging, blending extensive commercial and academic expertise. I have acquired hands-on experience at Broadcast Virtual and Bodymapp Ltd, complemented by research contributions at Queensland University of Technology and the University of Auckland. With a portfolio encompassing over 20 publications, I am committed to continuous learning and innovation in these fields.

SKills

3D SlicerAWS cloudBlenderC++Computer VisionDockerEEGFieldTripfMRIFSLLabVIEWMachine LearningMATLABMEGMeshLabMLFlowNeuroimagingNvidia DALIOpenMMLabOptunaPandasPybind11PythonROSScikit-LearnSensor data fusionShellSMPLSPMSuperviselyVTK
AI Research ScientistComputer Vision EngineerMachine Learning Engineer

Education

July 2010 - Aug 2014 PhD in Functional Neuroimaging at University of Otago
Aug 2008 - Oct 2009 Master's in Signal Processing at Linnaeus University
Feb 2022 - Jul 2007 Bachelor's in Electronics at Shahid Rajaee university

Experience

Sep 2022 - Present R&D Programmer at Broadcast Virtual Ltd

1) Image-based and Video-based semantic segmentation of the sports events (players and LED boards inside the stadiums):
● Images got extracted from various sports footage and sent to the external annotators.
● A Python pipeline was build using Albumentation and Nvidia.DALI libraries to preproccess and preaugment the training data.
● A training pipeline was made which included the following main parts:
• dataloader using Nvidia.DALI (GPU-based image & video processing),
• model maker: Pytorch and Timm libraries used for model loading and modifications,
• Optuna library was used for automatic hyperparameter search,
• MLFlow was used to log the model metrics/losses and trained weights.
● Trained models were converted to TRT format for real-time applications.
* So far two contracts have been secured using the above mention models for virtual LED advertisements in Japan and UK rugby events.
2. Visual Odometry for tracking lens cameras in sports events:
Visual Odometry is the backbone of the SLAM techniques. Following steps have been done to make Visual Odometry work for lens cameras inside stadiums:
● Conducting a literature review to identify the SoA techniques for keypoint extractors and keypoint matchers,
● Writing a Python pipeline to extract keypoints and match them and then optimize the camera parameters.
● Some of the algorithms such Dual SoftMax for matching was implemented in C++ for speed improvements.
* This project is still runnig and in the process of full conversion into C++.

Apr 2021 - Sep 2022 Computer Vision Research Engineer at Bodymapp Ltd

● Creating a Python pipeline to train a FCNN model for binary segmentation of the depth images for masking the human against the ambient,
● Creating a MATLAB based pipeline to:
• extract and preproccess the masked time-of-flight depth images,
• apply ICP to stitch depth frames to create 3D avatars.
● Creating a Python pipeline for optimizing the SMPL model to fit to avatars.
● Deploy the pipelines into the AWS EC2.
* A patent has been submitted for our work.

Feb 2018 - Feb 2021 Postdoctoral Research Fellow at Queensland University of Technology

● Engaged as a member of the medical robotics group, focusing on advancing technological solutions in surgical settings.
● Implemented 3D mapping techniques for surgical scenes in keyhole surgery, employing deep learning methodologies such as Fully Convolutional Neural Networks (FCNN) for depth
perception, semantic segmentation, and camera pose estimation during endoscopic procedures.
● Designed and constructed a stereo endoscope, including camera calibration to ensure accurate spatial perception and depth measurement.
● Contributed to the field by authoring reports and scientific articles, disseminating findings and advancements in medical robotic vision technologies.

Apr 2015 - Oct 2017 Postdoctoral Research Fellow at University of Auckland

Currently holding two concurrent part-time postdoctoral positions at the Physics Department and the School of Pharmacy, specializing in advanced neuroimaging data acquisition and fusion technologies.

● Responsible for recording neuroimaging data using EEG and fMRI modalities, ensuring high-quality data acquisition and synchronization.
● Implementing data fusion techniques to integrate simultaneous EEG and fMRI data, leveraging advanced signal processing algorithms.
● Developing real-time data recording systems for various sensors and devices such as photodiodes, spectrometers, DAQ systems, shutters, and laser sources to support experimental
setups.
● Establishing software repositories to manage and document lab equipment configurations, facilitating efficient operation and reproducibility of experiments.