Fans of IBM Z Hub

Fans of IBM Z Hub

Fans of IBM Z Hub

Join us and share the love of IBM Z with our global community!

 View Only

Our Experience Using IBM LinuxONE: From Datathon Participants to Project Winners

By Akshata Pandit posted Thu April 24, 2025 01:11 PM

  

Our Experience Using IBM LinuxONE:

From Datathon Participants to Project Winners 


Introduction: A Journey of Curiosity, Learning and Growth 

               Every once in a while, we come across opportunities that challenge us to go beyond the classroom, apply our knowledge in real-world scenarios, and evolve both as learners and creators. For us — Team Oasis from Global Academy of Technology, Bengaluru — the IBM Z Datathon was that opportunity. Comprising Akshata Pandit, Jhansi Prasad S, Charushila, Rachana N. S, and Dhanya Hegde, our team was drawn to the Datathon by our shared passion for technology and a desire to create something meaningful. From the beginning, we knew this wouldn’t be just another competition. It was going to be a journey of exploration, a dive into enterprise-level platforms, and a chance to create a real solution for a very real problem. 


                    We entered the Datathon with open minds and high hopes, and what unfolded was beyond what we expected — not only did we build a working prototype, but we also won the challenge and gained valuable mentorship from IBM. In this blog, we want to walk you through our experience, our project, and how IBM LinuxONE played a crucial role in helping us scale our ambitions. 


The Challenge We Took On: Protecting Lakes from Encroachment 


                   Living in a rapidly growing city like Bengaluru, we’ve all seen the impact of urbanization — and one of the most heartbreaking consequences has been the loss of our beautiful lakes. Encroachment, pollution, and negligence have made many of them vanish. That’s what inspired us to take on a project aimed at monitoring and preventing lake encroachment. 
Our goal was to build a solution that could take satellite and drone images of lakes and track changes in their boundaries over time. By calculating the area and identifying reductions or intrusions, we could offer a system that flags potential encroachments. It was an ambitious idea, and we knew it would require handling massive data, perform high-accuracy image analysis, and ensuring real-time responsiveness. 


Stepping into the World of IBM LinuxONE and L1CC 


                   At the start of the Datathon, we were introduced to IBM’s LinuxONE through a workshop, and from that moment, everything changed. The platform, which seemed distant and complex at first, turned out to be one of the most developer-friendly and powerful tools we’ve used. IBM LinuxONE Community Cloud (L1CC) provided us access to enterprise-class hardware through an Ubuntu-based Linux environment. It came pre-configured with Docker, and with the support of AI/ML libraries like TensorFlow, PyTorch, and Scikit-learn, we felt right at home. This wasn’t a restricted, rigid mainframe — it was an open, flexible, and modern development environment. What amazed us most was the performance. Unlike our local machines, which would choke under the weight of high-resolution images and deep learning models, IBM LinuxONE handled it all with ease. Whether it was running batch scripts for image processing or training CNNs for lake detection, L1CC performed smoothly and swiftly. 


Developing the Solution: Our Tech Stack in Action 


                     We began by collecting publicly available satellite and drone imagery of lakes. These images were often raw and noisy, so our first step was preprocessing using OpenCV — adjusting contrast, filtering out noise, and resizing them for uniformity. We then moved on to model training. We used Convolutional Neural Networks (CNNs), built in TensorFlow and Keras, to detect and classify water bodies and surrounding land areas. Some of our images showed significant encroachment, while others were clean — this imbalance in data 
prompted us to apply augmentation techniques like flipping, rotating, and scaling to improve generalization. Training the model locally would have taken hours, maybe days. But with LinuxONE, our models were trained in just minutes. It felt like having a supercomputer at our fingertips. We also built Python scripts to calculate the area of the lakes from the segmented images, which would then be used to detect and highlight reductions over time. 


The Real Magic: Deployment on IBM LinuxONE 


                 Once we had a model, we were proud of, it was time to deploy — and that’s where IBM LinuxONE truly shone. We containerized our model using Docker and moved it to the L1CC instance. The transition was effortless, thanks to the platform’s support for open standards and containerization. We did run into a hiccup where our model gave strange results on L1CC. After some investigation, we realized it was due to compatibility differences with the architecture, which we later solved by using updated PyTorch versions better suited for big-endian systems. Once deployed, the model responded with near-zero latency. We could input an image and get the analysis result almost instantly. This kind of performance is critical when working with large data in real-time, and it convinced us that our solution could be scaled and used in real-world deployments. 

                                                                                                                                                    
                                                                                                                                                       Architecture of the project 

Ongoing Mentorship and Project Evolution 


                     Winning the Datathon was an incredible milestone, but it was just the beginning. With mentorship from IBM, we began exploring how to make our project even more robust. We started thinking not just as students building a prototype, but as engineers creating a production-grade solution. Under guidance, we began wrapping our model into a REST API using Flask, so that other systems could interact with it remotely. We also looked into integrating dashboards and database support for historical tracking and visualization. With each step, the flexibility and scalability of IBM LinuxONE made things easier, not harder. The most exciting part is that we’re still learning, improving, and building — and IBM LinuxONE continues to be the platform enabling us to do so. Looking Back and Ahead: What IBM LinuxONE Meant to Us When we first heard about IBM Z and IBM LinuxONE, we imagined something abstract and reserved for large enterprises. But after this experience, we’ve come to see it as a welcoming, powerful, and future-ready environment where ideas can come to life. IBM LinuxONE is not just a cloud platform. It's a foundation built for security, performance, and scale. From the ease of setup to the speed of inference, everything about it empowered us to think bigger. We found that enterprise infrastructure doesn’t have to be complicated — when paired with open tools and great mentorship, it becomes a launchpad. This journey has taught us how impactful technology can be when it’s accessible, and how much can be achieved when the right tools meet the right intent. We’re proud of what we built, and even more excited about where it could go next


— Team Oasis 

0 comments
12 views

Permalink