Menu
Students Corner
Summer Internship
COE AI internship program is aimed at providing an opportunity for students for getting work experience in a rich and diverse research environment.
The summer internship program is a full-time, 6-8 weeks duration program. The internship is offered in the areas of:
- Machine Learning/ Deep Learning
- Image Processing
- Natural Language processing
- Control Systems
- Robotics
- Data Analytics
- Quantum Computing
Shortlisted students will have interviews on July 20th. The internship is set to begin on the 24th. Internees would receive certificates after successful completion of the internship program.
MS Thesis Topics
| Topics | Supervisor | Required Skills |
|---|---|---|
|
Heterogeneous health data Integration for AI Abstract: In order to promote AI analytics on health data, it is important to create a pathway from raw data to AI. Health data is available in heterogeneous formats from a variety of sources. A challenging task is to acquire and pre-process data in a standardized form while preserving privacy constraints associated with health data. |
Dr. Fatima Khalique |
Understanding of Health Data Interoperability Standards, Python/Java Programming Language |
|
Characterizing public attitude towards vaccination Abstract: Despite the availability of vaccines and vaccination services, the refusal of vaccine administration in both developed and developing countries has been a growing area of interest due to its impact on public health. The public attitude towards vaccines can be characterized based on NLP methods applied to social data augmented by vaccinated population data. |
Dr. Fatima Khalique |
NLP, Python Programming Language |
|
Contact Tracing using IoT and AI Abstract: Contact tracing is an effective prevention tool for interventions in the context of communicable diseases. An effective contact tracing mechanism if employed thoroughly, can break the transmission chain of the virus that is spread by close contact. In this regard, various digital contact tracing processes have been explored using technologies including Bluetooth, Global Positioning System (GPS), Social graph, contact details, network-based API, mobile tracking data, card transaction data, and system physical address. The digital contact tracing process ensures near real-time and reduced latency. The data collected through these technologies can then be analyzed through AI tools for tracing recent contacts of an infected individual. |
Dr. Fatima Khalique |
IoT, Python Programming Language |
|
A Visual Hearing Enabler Using Lip to Speech Synthesis Abstract: Visual hearing refers to, understanding silent speech. Babies observe lip movements when they start learning, adults involuntarily tend to comprehend speech with lip reading in a high-noise environment. Visual hearing application is a lip-to-a-speech synthesizer that uses nonauditory sources i.e. lip movement to generate speech. This tool has many applications where auditory information of the speech is either very weak, corrupt or missing. It can be used for speech recovery from the noisy environment, speech generation for people having aphonia, and speech inpainting for missing or corrupted voice segments. The visual hearing tool would be based on deep learning approaches for synthesizing speech sequences from silent speech videos. |
Dr. Sumaira Kausar |
Machine Learning concepts, Deep learning concepts, python language for programing |
|
Furniture Audio Description to Furniture Image Translation Abstract: Furniture design starts from a design concept that is usually communicated verbally. This concept is then used to build a prototype that is approved and ultimately converted into actual furniture. The Furniture Audio Description to Furniture Image Translation system can facilitate the furniture designers to translate audio design concepts into a soft prototype. That would give an effective solution for design concept approval and prototype visualization. The system would be developed using a deep learning-based translation model. |
Dr. Sumaira Kausar |
Machine Learning concepts, Deep Learning concepts, Python Language for Programing |
|
Ordinance Categorization and Semantic Modelling Abstract: The work is based on two techniques. Initially, the Copyright Act and Trademark Ordinance should be Categorized and secondly, related cases will be categorized. Semantic Identify the similarity of user query with the categorized data. The data set should develop and analyze the dataset for accuracy. |
Dr. Muhammad Asfand-e-Yar |
Natural Language Processing concepts, Machine Learning concepts |
|
Clinical Drug interactions with associative Events Abstract: The work is based on the databank of drugs/salt. The interaction of drugs with food items and personal activities, according to the usage of medicine during illness. Implementing an application for the local community by integrating the salt with drug names used in the majority of medical stores or local pharmaceutical companies. |
Dr. Muhammad Asfand-e-Yar |
Natural Language Processing concepts, Machine Learning concepts |
|
Implementation of AI-Based Approach for Disturbance Estimation of a MIMO Nonlinear System Abstract: Diagnosis of early and later age balance disorder caused by the central nervous system. Due to this imbalance falls can be prevented as a natural consequence of a variety of balance disorders. This research focuses on postural stability and sit-to-stand (sts) movement execution in the presence of feedback delays and noisy joint sensors by using a nonlinear controller and mathematical model of the musculoskeletal system. Nonlinear feedback extended high gain observer (ehgo) is designed to regulate and estimate the movement of a biomechanical model for postural stability and fall prevention. Moreover, the traditional machine learning (ml) approach is used, in which input parameters are defined (noise, disturbance, torque) and then, implement feature extraction on input parameters. Although, the regression model is designed by neural networks, support vector machine regression, decision tree regression, and random forest regression which predicts output. The purpose of this research is to create a neuromechanical model with a nonlinear and stable extended high-gain observer-based controller to govern postural and biomechanical sit-to-stand actions. The simulations will be performed in Matlab, Simulink, and python. Then, the results will show that our suggested approach will minimize the risk of falling and improve the exoskeleton design. |
Ms. Nadia Imran |
Matlab/ Simulink, Python (google colab/ spyder/ jupiter notebook/ pycharm), Latex. |
|
Detection of Bots in Social Media Abstract: Social Media Bots are kinds of attacks that harm many domains related to health, economy, politics, public opinions and many more. In the recent past, many techniques are proposed to detect these bots mostly targeting Twitter. The proposed project will explore the problem on other social media platforms. |
Dr. Saba Mahmood |
Python Programming Language |
|
Detection of Sybil Attacks in Online Social Networks Abstract: The immense popularity of Social Media has made it vulnerable to many attacks like the Sybil attack. In this attack, numerous fake identities are created, either sending friend requests or cyberbullying. The project will explore techniques to detect the Sybils using graph-based approaches. |
Dr. Saba Mahmood |
Python Programming Language |
MS Thesis Topics
| Topics | Supervisor | Required Skills |
|---|---|---|
|
Heterogeneous health data Integration for AI Abstract: In order to promote AI analytics on health data, it is important to create a pathway from raw data to AI. Health data is available in heterogeneous formats from a variety of sources. A challenging task is to acquire and pre-process data in a standardized form while preserving privacy constraints associated with health data. |
Dr. Fatima Khalique |
Understanding of Health Data Interoperability Standards, Python/Java Programming Language |
|
Characterizing public attitude towards vaccination Abstract: Despite the availability of vaccines and vaccination services, the refusal of vaccine administration in both developed and developing countries has been a growing area of interest due to its impact on public health. The public attitude towards vaccines can be characterized based on NLP methods applied to social data augmented by vaccinated population data. |
Dr. Fatima Khalique |
NLP, Python Programming Language |
|
Contact Tracing using IoT and AI Abstract: Contact tracing is an effective prevention tool for interventions in the context of communicable diseases. An effective contact tracing mechanism if employed thoroughly, can break the transmission chain of the virus that is spread by close contact. In this regard, various digital contact tracing processes have been explored using technologies including Bluetooth, Global Positioning System (GPS), Social graph, contact details, network-based API, mobile tracking data, card transaction data, and system physical address. The digital contact tracing process ensures near real-time and reduced latency. The data collected through these technologies can then be analyzed through AI tools for tracing recent contacts of an infected individual. |
Dr. Fatima Khalique |
IoT, Python Programming Language |
|
A Visual Hearing Enabler Using Lip to Speech Synthesis Abstract: Visual hearing refers to, understanding silent speech. Babies observe lip movements when they start learning, adults involuntarily tend to comprehend speech with lip reading in a high-noise environment. Visual hearing application is a lip-to-a-speech synthesizer that uses nonauditory sources i.e. lip movement to generate speech. This tool has many applications where auditory information of the speech is either very weak, corrupt or missing. It can be used for speech recovery from the noisy environment, speech generation for people having aphonia, and speech inpainting for missing or corrupted voice segments. The visual hearing tool would be based on deep learning approaches for synthesizing speech sequences from silent speech videos. |
Dr. Sumaira Kausar |
Machine Learning concepts, Deep learning concepts, python language for programing |
|
Furniture Audio Description to Furniture Image Translation Abstract: Furniture design starts from a design concept that is usually communicated verbally. This concept is then used to build a prototype that is approved and ultimately converted into actual furniture. The Furniture Audio Description to Furniture Image Translation system can facilitate the furniture designers to translate audio design concepts into a soft prototype. That would give an effective solution for design concept approval and prototype visualization. The system would be developed using a deep learning-based translation model. |
Dr. Sumaira Kausar |
Machine Learning concepts, Deep Learning concepts, Python Language for Programing |
|
Ordinance Categorization and Semantic Modelling Abstract: The work is based on two techniques. Initially, the Copyright Act and Trademark Ordinance should be Categorized and secondly, related cases will be categorized. Semantic Identify the similarity of user query with the categorized data. The data set should develop and analyze the dataset for accuracy. |
Dr. Muhammad Asfand-e-Yar |
Natural Language Processing concepts, Machine Learning concepts |
|
Clinical Drug interactions with associative Events Abstract: The work is based on the databank of drugs/salt. The interaction of drugs with food items and personal activities, according to the usage of medicine during illness. Implementing an application for the local community by integrating the salt with drug names used in the majority of medical stores or local pharmaceutical companies. |
Dr. Muhammad Asfand-e-Yar |
Natural Language Processing concepts, Machine Learning concepts |
|
Implementation of AI-Based Approach for Disturbance Estimation of a MIMO Nonlinear System Abstract: Diagnosis of early and later age balance disorder caused by the central nervous system. Due to this imbalance falls can be prevented as a natural consequence of a variety of balance disorders. This research focuses on postural stability and sit-to-stand (sts) movement execution in the presence of feedback delays and noisy joint sensors by using a nonlinear controller and mathematical model of the musculoskeletal system. Nonlinear feedback extended high gain observer (ehgo) is designed to regulate and estimate the movement of a biomechanical model for postural stability and fall prevention. Moreover, the traditional machine learning (ml) approach is used, in which input parameters are defined (noise, disturbance, torque) and then, implement feature extraction on input parameters. Although, the regression model is designed by neural networks, support vector machine regression, decision tree regression, and random forest regression which predicts output. The purpose of this research is to create a neuromechanical model with a nonlinear and stable extended high-gain observer-based controller to govern postural and biomechanical sit-to-stand actions. The simulations will be performed in Matlab, Simulink, and python. Then, the results will show that our suggested approach will minimize the risk of falling and improve the exoskeleton design. |
Ms. Nadia Imran |
Matlab/ Simulink, Python (google colab/ spyder/ jupiter notebook/ pycharm), Latex. |
|
Detection of Bots in Social Media Abstract: Social Media Bots are kinds of attacks that harm many domains related to health, economy, politics, public opinions and many more. In the recent past, many techniques are proposed to detect these bots mostly targeting Twitter. The proposed project will explore the problem on other social media platforms. |
Dr. Saba Mahmood |
Python Programming Language |
|
Detection of Sybil Attacks in Online Social Networks Abstract: The immense popularity of Social Media has made it vulnerable to many attacks like the Sybil attack. In this attack, numerous fake identities are created, either sending friend requests or cyberbullying. The project will explore techniques to detect the Sybils using graph-based approaches. |
Dr. Saba Mahmood |
Python Programming Language |
|
Abstractive Text Summarization for Roman Urdu Abstract: Text Summarization is the process of reducing the overall length of a given text while retaining the main information present in it. In this regard a lot of work has been done for the English language however for Urdu there do not exist a lot of references. Generally, the techniques can be divided into two categories; (1) Extractive i.e. picking representative sentences directly from the text and using them as a summarization. (2) Abstractive i.e. generating new text from existing sentences to present a summary. |
Dr. Momina Moetesum |
Python, NLTK |
|
Pashto and Punjabi Speech Recognition System with focus on agriculture related vocabulary Abstract: Automatic speech recognition (ASR) is the process of understanding speech and converting it to text. Therefore, ASRs are also known as speech-to-text converters. Recent developments in ASRs have led to extremely large models that have been trained in several languages. Such methods focus mostly on acoustic models and do not rely on language models for the generation of text. The advantage of such methods is that they can recognize words that the system has not already seen. The focus of this work is to explore such a speech recognition system called XLSR (wav2vec). It is a self-supervised cross-lingual speech representation library, as per the description of the authors. The existing model has been trained for more than 50 languages. However, the number of samples for each language is not the same. Some low-resource languages are part of the system but require more data to be finetuned i.e. perform better. |
Dr. Imran Siddiqi |
Python |
|
Pest Detection in Images Abstract: Farmers are often unable to identify the pests/diseases attacking their crops and hence require assistance. Assistance in the form of field agents may take days to reach the farmers and this delay may result in loss of time and hence affect the yield of the crop. We are interested in developing an automated solution where the farmer can follow simple instructions to take pictures of the pests/disease attacking their crops and get information about the disease/pest. Such a solution may be achieved by developing a smartphone application. The farmer can take pictures of the pests/disease using the application and the application can identify the pests for them and then present remedial measure es to the farmer. |
Dr. Momina Moetesum |
Python & Associated Libraries |
|
Land cover and crop classification in satellite images Abstract: The utilization of satellite data is becoming more and more important. It is an effective tool for monitoring on-ground situations with high frequency and over large areas. The current floods/rain devastation that has affected Pakistan is being monitored using satellite data. Satellite data is also important in identifying the types of crops that are cultivated, the health of the crop and to identify the number of nutrients in the soil. This helps in improving the production per hectare i.e. the yield of a crop. The focus of the proposed work is to first perform land cover classification by identifying crop fields in a satellite image. In the second step, the focus is to identify the area in which a particular crop is cultivated based on its type. For the work, we shall consider using open access Sentinel 2 data. |
Dr. Imran Siddiqi |
Python |
|
Crop Yield Estimation Abstract: Farmers are interested in estimating the yield of a crop. This helps them identify the amount of money they can make at the time of cultivation and hence plan the next crop cycle. This information is also important for contract farmers as this helps them protect themselves from underproduction, which may force them to buy crops from the market. This also helps government agencies determine the production of a crop and identify any shortages or excess in general requirements. Most articles in the literature use soil information, temperature and humidity as input variables. Soil information may contain a soil map (nutrients in soil, type of soil and location), soil type, pH value and area of production. Crop information i.e. weight, growth, variety, density and leaf area index is sometimes combined with soil information along with humidity and mean rainfall to predict the yield of a crop. |
Dr. Imran Siddiqi |
Python |
Final Year Projects
| Topics | Supervisor | Required Skills |
|---|---|---|
|
Apparel Design Application Abstract: Preparing apparel is a multistage process, from conceptualization to the final product. Designers prefer to have a prototype of the final product for approval and testing. This can help to save their time, effort and cost. The Apparel Design Application will enable the designer to convert an apparel sketch into rendered apparel. This would be a time, effort and cost-effective solution to developing a prototype. |
Dr. Sumaira Kausar |
Machine Learning concepts, Deep learning concepts, Generative Adversarial Networks concepts, Python Programming Language |
|
Virtual Physiotherapist Abstract: Virtual reality has shown promising results in the rehabilitation of many physical and psychological conditions. The virtual physiotherapist would be a virtual reality-based rehabilitation tool that would help individuals with cervical and upper limb conditions. This tool would focus on physiotherapy sessions for individuals having work-related neck and upper limbs musculoskeletal disorders. |
Dr. Sumaira Kausar |
Virtual Reality, Unity 3D |
|
Smart Home Budget System Abstract: The project is for the fiscal year budget analyzer. A database (RDB) is to be designed, based on the information provided by the supervisor. The Mobile application will get a daily expense report for a user, which will propagate next month’s budget expense accordingly. |
Dr. Muhammad Asfand-e-Yar |
Software Tools: React JS and Node JS, Database: MySQL |
|
Integration of Exterior and Interior Home Décor Abstract: The project is based on the combination of two projects that is Exterior and Interior Home décor. The Interior décor is working on, Exterior décor will be designed and integrated with already interior design. |
Dr. Muhammad Asfand-e-Yar |
Software Tools: Blender (for models and objects) and React JS (Front application), VR/AR Programming |
|
Roman Urdu and English Chat Bot for Property Deals Abstract: The project is based on the chatbot NLP techniques. A set of FAQs will be provided to students. An automated system is to be designed for local communities to provide facilities for querying in Real Estate. |
Dr. Muhammad Asfand-e-Yar |
Software Tools: Python, Node JS and PHP, Database: MySQL, Dataset: CSV files |
|
Wearable Knee Recovery Device Abstract: Total knee replacement (tkr) surgery is the most common orthopedic procedure globally spearheaded by aging societies, and the procedure is commonly followed by a long, unengaging, often (perceived as) burdensome rehabilitation period. The rehabilitation field has barely seen changes in the past 50 years and the advent of aging populations has only proven to worsen the already overburdened healthcare workforce. The need for efficient, highly efficacious treatment is imperative and we believe the recent advancement of wearable IoT technology, now at the forefront of personal health monitoring, is the key to breaking the field. Currently, knee problems are becoming more common and increasing more rapidly as many people suffer from permanent disability due to joint problems, orthopedics issues, muscle injuries and accidents. To overcome this rapid problem, we are designing a wearable recovery device that will provide additional torque for the knee joints by processing the signals and the angle of the knee. Using this system, it is expected that the number of visits to hospitals as well as the number of interventions can be reduced, as well as the associated cost of care. Moreover, our design will reduce the stress on joints and provide easy movement while mounting stairs and walking. |
Ms. Nadia Imran |
Microcontroller or Embedded Basics, C/Python Language Programming with emphasis on real-time concepts, Embedded Software Programming/Debugging skills, Machine Learning, Robotic Control Systems, Hardware Design |
Career
No jobs found