TUKL-NUST Research & Development Center

Research projects 

TeleDoc: A Portable Device for Remote Diagnostic Screening of Patients

A portable, user-friendly android application integrated with an artificially intelligent expert system developed for the early detection of Dengue and prediction of progression of disease from one stage to another.

Dr. Faisal Shafait, Nyma Altaf, Ramsha Waseem


An Automatic Real Time Vehicle Detection, Identification and Registration Plate Recognition System

Development of hardware based Automatic Real Time Vehicle Detection, Identification and Registration Plate Recognition System to improve road security in Pakistan.

Dr Faisal Shafait, Khurram javed, Amna Masood, Imran Malik, Kuldeep Langhani, Noor Ul Sehr Zia, Talha Paracha, Zaheer, Zainab Tareen


Fish Biodiversity Estimation by Low-Cost Non-Destructive Video Based Sampling (FIBEVID)

The objective of FIBEVID is to utilize computer vision and deep learning techniques to survey fresh water fish species. FIBEVID aims at producing a semi-automatic surveying tool that minimizes human interaction and maximizes accuracy based on uncertainty querying.

Dr. Faisal Shafait, Syed Hassaan Tauqeer, Rafay Ahmed Khan, Abdul Hannan Khan, Amal Mubashar


Deep Learning based Fish Detection, Counting and Classification

Monterey Bay Aquarium Research Institute(MBARI) is an oceanographic research center based in California. They have developed a software named Automated Video Event Detection(AVEDAC) for underwater fish detection. We at TUKL are collaborating with them to replace the fish detection and classification modules of AVEDAC deep learning based model.

Dr. Faisal Shafait, Talha Rehman, Salman Maqbool, Shah Rukh Qasim


Fish detection from underwater videos

Detect fish from videos through motion cues obtained using optical flow.

Dr. Faisal Shafait,Dr. Ahmad Salman, Salman Maqbool, Talha Rehman


Automatic Signature Segmentation in Hyperspectral Document Images

Proposed method is based on part based key detection and does not use any color or structure information for segmentation. It will make use of hyper spectral responses to segment the signature in overlapping and non overlapping situations.

Dr. Muhammad Imran Malik, Umair Munir


Text-line Feature Learning Using RBM & Stacked Auto-encoders

Learning the features using whole text lines. To compare the OCR results for features learnt from RBM and from Stacked Auto-encoders. LSTM networks are used for training OCR using the features learnt on UPTI database.

Dr. Faisal Shafait, Aqsa Ahmed


Unconstrained handwriting recognition

Recognizing unconstrained handwriting is a complex problem. Variations between writers, overlapping words, changing skew, changing slant and other properties lead to bad recognition rates on models designed for printed text. We are working on better recognition rates using the context aware properties of Bi-LSTM after extracting multiple features from the training data.

Dr. Faisal Shafait, Muhammad Ferjad Naeem, Tooba Imtiaz


Recognition of Blurred Documents

Motion and Gaussian blur removal from the document images using the deep learning approach. The LSTM (long-short term memory network) is employed for the blur removal.

Dr. Faisal Shafait, Fallak Asad


Exploration and implementation of document classification methods

Document image classification for automatically assigning an unseen document image to one of the several categories depending on its content which also provides the better results with blurred, distorted or partially-visible forms of images.

Dr. Faisal Shafait, Fallak Asad


Document Understanding: Using elastic matching to extract form data

Proposed method is based on using elastic matching techniques to extract filled form data from raster forms' OCR and handwriting recognition result. We have started from the very old techniques but our aim is to solve the problem using machine learning.

Dr. Faisal Shafait, Shah Rukh Qasim, Talha Rehman, Tooba Imtiaz, Muhammad Ferjad Naeem


Document Understanding Using Long Short Term Memory (LSTM) Neural Network

Analysis of the logical layout of documents not only enables an automatic conversion into a semantically marked up electronic model but also affirms option for maintaining higher level functionality.

Dr. Faisal Shafait, Muhammad Kamran, Shah Rukh Qasim


Business Card Reading and Text extraction

Extraction of information from business cards of Asian languages using image processing techniques.

Dr. Faisal Shafait, Fallak Asad


Detection of German Language Text and Music Symbols from a Music Chart

A music chart contains not only text but also music symbols. Extraction of this information is a challenging task. by using image processing techniques extract text and music symbols from music chart.

Dr. Faisal Shafait, Muhammad Mohsin Ghaffar


FPGA Based Video Processing

FPGAs are great fits for video and image processing applications, capable of performing computations at significantly faster rates than GPUs. The object of this project is to implement image processing and machine learning algorithms on FPGAs to be used independently in the hardware.

Dr. Faisal Shafait, Mohsin Ghaffar, Nauman Mustafa, Amna Masood, Zainab Tareen




The objective of this project is to develop intelligent tools for real-time monitoring and analysis of SDN log data. The proposed system will ideally enable PROACTIVE monitoring of SDN infrastructure by leveraging state-of-the-art techniques in the areas of machine learning and natural language processing.

Dr. Faisal Shafait, Ahad Mahmood, Muhammad Moazzam Khan, Rana Awais


Business Analytics for Industrial IOT using LoRa Mesh (BA-iLM)

Business Analytics for Industrial IOT using LoRa Mesh (BA-iLM) is a funded project by Unilever which uses LoRa Technology and Sensor Networks for data acquisition from different parts of Unilever's Wall's Factory. Data gathered from the factory is then integrated and analysed using concepts of Business Intelligence and Data Warehousing in order to produce better visualizations and reports to facilitate better decision-making.

Dr. Faisal Shafait, Syed Ahmed Hussain Naqvi


Computer Vision Based UAV Landing System

Development of algorithms to allow a UAV to automatically detect and land on a runway using input from its camera.

Dr. Faisal Shafait, Haider Ali Agha, Taimoor Tariq, Ayesha Yaqub


Smart and portable device for accurate diagnostic screening.

Development of an intellegent, portable and cost effective diagnostic screening device for common and epidemic diseases.

Dr. Faisal Shafait, Sohaib Nasir, Abdul Rahman bin Saad, Abdullah Anjum


Clothing Retrieval System based on Convolutional Neural Network

The system will let the user query the clothing data set and retrieve garments based on visual similarity.

Dr. Faisal Shafait, Ayesha Khalid, Fatima Arshad


Underwater Imaging for Fish Fauna Estimation

This project aims at developing a non-invasive software system to detect and calculate the statistics of fish species in open waters of Pakistan. This software will be aided by a hardware that will help us in collecting fish data. This hardware will be developed to record video data of underwater environment containing fish species under observation.

Dr. Faisal Shafait, Abdul Hannan Khan, Amal Mubashir


Cargar: The Next Generation Motorbike Taxi

To create Pakistan's first and only motorbike taxi hailing application. To incorporate cutting edge features like speech recognition and understanding and build a great user experience.

Dr. Faisal Shafait, Muddassir Ahmed Khan