Artificial Intelligence (AI)
Discuss current events in AI and technological innovations with Intel® employees
676 Discussions

Developers Built Impactful Solutions at the Intel® AI Hackathon (IEEE INDICON 2024)

Nikita_Shiledarbaxi
0 0 969

INDICON 2024, the flagship conference of the IEEE India Council was organized at the Indian Institute of Technology (IIT), Kharagpur (ranked as #5 among the top national engineering institutes declared by the Govt. of India in 2024)[1] in December'24. Intel, in collaboration with IEEE and IIT Kharagpur, marked its remarkable presence at the event by hosting an Intel® AI Hackathon. The hackathon was an AI readiness-focused initiative for next-gen developers, AI enthusiasts and technocrats from higher educational institutions in India which provided them with an opportunity to create and showcase AI innovations.

 

The event was led by Jayaraman Mahalingam (JAY) (an AI/oneAPI Academia Tech Program Lead from the Software Ecosystem Enablement (SEE) group at Intel) in association with executives from Beahead Pvt. Ltd.* (Intel’s partner).

Here are the event highlights:

  • 1.4K+ students forming 468 project teams, representing 199 institutions and 25 states in India registered for the event.
  • 122 project submissions were received, with the participants coming from 76 different Indian institutions and 21 distinct states.
  • Hackathon project themes: Education, Healthcare, manufacturing, Safety and Open Innovation.
  • The hackathon comprised of multiple phases beginning with registration process in November followed by selection of top 30 teams in early December, 3 levels of live pitch by the selected project teams and winners’ announcement during the INDICON event towards end of December. Experts from Intel also provided the participants with online training on oneAPI-powered developer resources.

 

 

“The hackathon generated unmatched enthusiasm among the student community and received 1200+ individual registrations and nearly 500 team registrations (across 5 tracks) within a registration window of 2 weeks.

 

The Intel AI Hackathon received approx. 120 eligible (complete) submissions, 30 were shortlisted for pitching (online) at the national semi-final, and the top 10 teams were invited to present in person at IEEE IndiCon.

 

We take this opportunity to thank the entire team of Intel and its partners led by Mr. Jayaraman Mahalingam and Pradipta Sharma, the IEEE YP Affinity of IIT Kharagpur represented by Mr. Aditya Kameswara Rao Nandula, Mr. Partha Acharya, and IEEE UPP Student volunteers, the competition judges, and the IEEE IndiCon 2024 Team for their support, encouragement, and dedication in making the event a grand success. We thank all the individuals and team members for their enthusiastic participation and involvement.”

 

-Dr. Subhamoy Mandal, Assistant Professor at IIT Kharagpur

 

 

Read further to learn about the winning projects at the AI hackathon.

 

Winning Projects at the Intel® AI Hackathon

The competing projects were judged based on a variety of criteria such as degree of alignment of the solution with the hackathon theme and the problem statement of the chosen track, impact and innovation, readiness of the solution, and use of Intel® technologies in developing the project.

Following were the top 4 projects that won the hackathon:

 

1. Project ‘Advanced Health Monitoring System Using AI & ML Techniques for Predictive Maintenance’ based on the ‘Open Innovation; theme won the 1st-prize.

 

The health monitoring system employs various AI/ML techniques and uses real-time data from sensors that monitor temperature, humidity, acceleration and distance for continuously assessing the health of equipment. It provides a proactive maintenance solution that minimizes downtime and enhances operational reliability in various industrial applications.

 

The team from IIT Kharagpur utilized Intel® oneAPI Data Analytics Library (oneDAL) for efficient data pre-processing, feature extraction, and ML model optimization. The library significantly helped them improve the performance of their data processing pipeline, enabling faster computations and better scalability.

 

2. Project ‘BloomScore’ based on ‘Education’ theme secured the 1st runner-up position. The team ‘Alchamps’ came from IIT Kharagpur.

 

BloomScore helps educators design and assess evaluations according to Bloom's Cognitive Taxonomy levels. By employing a fine-tuned DistilBERT model, it classifies questions or assignments based on their cognitive complexity, aiding educators in determining the cognitive level of their assessments. The platform allows educators to create tailored assessments by uploading topic PDFs and generating questions that match the specified Bloom's levels. This ensures that assessments are well-rounded, fostering deeper learning and critical thinking skills.

 

The team leveraged PyTorch* Optimizations from Intel in conjunction with GPU available on the Intel® Tiber™ AI Cloud platform for faster training and fine-tuning of Hugging Face* DistilBERT and Flan 5 models. They also utilized the Intel Neural Compressor tool for reducing the size of the Flan 5 model by half. The team found Intel’s technologies to be significantly useful for enhancing their project’s efficiency, scalability and performance.

 

Refer to the project demo for more details.

 

3. Project ‘CardioLens’ based on the ‘Healthcare’ theme and developed by the team ‘Three of Hearts’ from the Rajiv Gandhi University of Knowledge and Technology was ranked as the 2nd runner-up.

 

The project leverages AI for automated echo-cardiogram analysis. Using the Intel® Extension for Transformers for efficient inference, the team offers a solution that provides accurate and timely heart disease detection. The web interface of the project enables users to upload heart videos, generating segmentation masks and disease predictions with Intel® optimized model inference techniques. In collaboration with healthcare professionals, the solution alleviates key challenges faced by clinicians and students, automating manual tasks and enhancing diagnostic workflows with AI models.

 

The team utilized Intel® Extension for PyTorch for model optimizations on Intel CPUs and GPUs. They also benefited from the Intel® oneAPI Deep Neural Network (oneDNN) library for faster model training and inference through hardware acceleration and kernel optimizations. They also achieved 50% reduction in inference time using the Intel® Distribution of OpenVINO™ Toolkit as compared to that with the stock PyTorch.

 

Check out the project demo for more details.

 

4. Project ‘AI enabled Prosthetic Design Using OpenVINO Toolkit’ built by team ‘Horizon’ from the Heritage Institute of Technology and based on the ‘Healthcare’ theme also secured the 2nd runner up position at the hackathon. By leveraging advanced deep learning models, the solution processes CT scan images to identify and segment relevant anatomical structures with precision. The output is a high-quality STL file, ready for 3D printing, ensuring personalized, accurate and customizable prosthetic manufacturing. This project addresses key challenges in prosthetic design by reducing costs, improving accessibility, and automating traditionally complex workflows.   With its user-friendly interface and optimized performance, the project bridges the gap between cutting-edge AI technology and practical healthcare applications, offering transformative impact globally.

 

The team employed Intel Distribution of OpenVINO Toolkit for optimizing U-Net segmentation model using Neural Network Compression Framework (NNCF) for INT8 quantization and OpenVINO Runtime for dynamic device allocation, achieving 79% faster inference. They used AI tools and model optimized from Intel Tiber AI Cloud on AI PC powered by Intel® Core™ Ultra Processor, Intel® Data Center GPU Max Series and 4th Gen Intel® Xeon® Scalable Processor with Intel® Gaudi® AI accelerator, enabling real-time prosthetic design workflows with ~98% reduction in inference time.

Refer to the project demo for more details.

 

What’s Next?

We encourage you to explore our AI tools and frameworks and AI PC development kits. Sign up to Intel Tiber AI Cloud - try your hands on our optimized software and the latest accelerated hardware for high-performance AI and accelerated computing.

Also check out AI and HPC tools included in our oneAPI-powered software portfolio.

 

Additional Resources

 

[1] https://www.nirfindia.org/Rankings/2024/EngineeringRanking.html

About the Author
Technical Software Product Marketing Engineer, Intel