Advisors: Dr. Amara Tariq, Dr. Nazim Ashraf
Student Members: Ammar Saqib, Sarah Sajid, Sheikh Mahad Arif
In the era of automation and artificial intelligence, every aspect of life now requires tech-based solutions implemented on faster and more efficient methods; this is where deep learning thrives. However, a major roadblock for designing deep learning based solutions to any problem is the requirement of huge amount of labelled data for training. Domain Adaptation is designed to alleviate this problem by ensuring that a model trained over available labelled data of source domain generalizes well over unlabelled data of target domain.
Various techniques have been successfully applied to adapt deep learning based image classifiers from source to target domain data. To bring Pakistan into the race of advancements in deep learning, we explored domain adaptation techniques to design it for a much more complex image processing system, i.e., lane marking that involves instance segmentation for self-driving cars. Since it is crucial for lane marking systems to be able to adapt to different domains as traffic scenarios vary across countries, road types, etc., self-driving systems are in dire need of domain adaptation.
We developed a lane marking model that successfully generalizes from highway-traffic images to images of locally collected traffic data from Lahore, Pakistan. Our system shows that domain adaptation significantly improves the performance of lane marking system for unlabelled data of a target domain.
Dr. Saad Bin Saleem
Student Member: Rida Ayub
Nowadays, the robot supported education is becoming commonplace in various international schools in addition to the traditional methods of teaching. However, the usage of robot supported education is not common in the schools of Lahore, Pakistan. To fill this gap, we are introducing a shape drawing robot (Artie 3000) to some schools of Lahore to get feedback from students on the usability of this robot.
In this study, we teach students to draw geometric shapes using Artie 3000 robot; after that we ask students to draw shapes on their own. We observe students while they are drawing the shapes and ask questions from them about their experience of using the Artie 3000 robot. We are not limited to use the default shapes available in the Artie 3000 robot rather we have customized the robot to include some complex shapes in its database.
In our experiments, the students are evaluating the interface of Artie 3000 robot by drawing both default and new (customized) shapes. We are involving the students of elementary classes e.g. second, third and fourth grades in this first study. This is an ORIC (FCC) funded project.
Advisor: Mr. Muhammad Salman Chaudhry
Student Members: Ali Iftikhar, Abdullah Butt, Shayan Zafar
Pakistan has seen an exponential growth in tourism in the past few years which shows a promising potential for new investment in this industry. The project “Automated Binocular” was conceived exactly under the influence of this growing trend.
Sightseeing and looking for popular landmarks is a favorite activity of tourists visiting a travel destination. Therefore binoculars have been placed in famous tourist places for a long time by concerned authorities for tourists. The experience of a stereo optical view has proven to be much superior to a digital image so far, hence traditional binoculars are still interesting.
These binoculars however, are of not much use for tourists if they are not aware of where to look and for what. Our project provides the solution to this problem. The user is able to select a landmark from a number of options using a touchscreen and the binocular automatically aligns itself to view the selected landmark. This is achieved using Machine Learning, embedded systems and a motorized assembly. The prototype uses Google maps to locate the coordinates and align itself roughly based on its own coordinates. Then it aligns itself accurately using Machine Learning by recognizing the selected landmark using a camera. The prototype of the project has received considerable appreciation from observers.