Research and Development of Projects in Machine Learning and Computer Vision

Project: Automatic Number Plate Recognition System for Pakistani Vehicles
• Localization of standard and non-standard vehicle\'s license plate using YOLO model and Hough Line Transformation techniques
• Recognizing license plate number by applying image processing techniques and training a Convolutional Neural Network
• Creating pipeline for synchronized real time simulation of the localized & recognized deep learning models
• Designing the system’s Graphical User Interface in .NET Framework
• Programming Language Used: C++

Project: Secure Eye - A real time home surveillance system designed for person and vehicle classification (in collaboration with Nayatel Pvt. Limited)
• Creation of the synthetic data set through digital image filtering techniques
• Statistical evaluation of the deep learning model using data visualization tools & techniques
• Deployment of the model at different home premises
• Programming Language Used: Python

Project: Automatic Target Localization through Radars
• Extraction of the target data (digits, letters) from radar software using image processing techniques
• Applying Optical Character Recognition to extracted target data
• Successful Integration of PTZ cameras with the recognized target data values
• Programming Language Used: Python

Master Thesis: Mapping Sensor Data To Two Dimensional Depth Loss Profile Using Deep Learning Based Systems.

• Creation, analysis and selection of different real & artificially created Magnetic Flux Leakage (MFL) sensor data sets for train, validation and test purpose
• Successful generation of depth profile images form MFL images by employing Conditional Generative Adversarial Network (cGAN) for solving Image-to-Image translation tasks
• Accomplished best statistical performance on test data set by varying the architecture of cGAN
• Successful prediction of depth profile (wall loss) from MFL image by training a Convolutional Neural Network (CNN)
• Comparison between Conditional GAN and CNN through data visualization tools and techniques
• Tool used: TensorFlow framework in Python environment

Project Title: Identification of Van Gogh Painting Images and Their Forgeries using Deep Learning

• Classification of Van Gogh’s paintings among RGB density-normalized images
• Implementing a CNN using pre-trained model ‘VGG-F’ - Transfer Learning Concept
• Applying dimensionality reduction techniques to represent features in best dimension
• Classification using Support Vector Machine and Neural Pattern Recognition Tool Box classifiers
• Classification results comparison between feature engineering and feature learning approach
• Tool used: MATLAB

Department of Transmission, VoIP & Optical Fiber Cable. 1. Worked on transmission system of Huawei & introduction to Huawei switching system. 2. Studied the basic setup of telephone network, the protocols used in NGN and test equipment used in optical fibers

Project on the design, development and automation of Computer Numerical Control Wood Router. 1. Designed the prototype of wood router on Solid Works. 2. Software and electronic development using Computer Aided Design (CAD) & Computer Aided Machine (CAM) software to design patterns. 3. Automation of mechanism using stepper motors running through TB6560 micro controller