24/01/2025
Meeting #1
In my first meeting with my supervisor, Dr. Sumitra Kotipalli, Discussed the project idea and initial proposal. The session focused on outlining the core objectives of the project, which include real-time cricket ball tracking using OpenCV and a custom dataset. The supervisor provided initial feedback on refining the scope to ensure the project remains practical within the given timeframe. We also discussed potential challenges, such as handling different ball speeds, varying lighting conditions, and ensuring accurate detection regardless of background noise. Additionally, there was a brief discussion on the importance of maintaining a structured approach for dataset creation, as it will play a critical role in achieving reliable results.
27/01/2025
Meeting #2
Finalized the project and discussed the final proposal. In this meeting, the supervisor reviewed the revised project proposal, which now included a well-defined methodology and expected outcomes. The discussion covered technical aspects, including the choice of tools (Flutter for app development, Firebase for the database, and OpenCV for ball detection). The supervisor provided insights on structuring the dataset collection process and how to best analyze the results during testing. Furthermore, we touched upon project scheduling, ensuring that each phase—from data collection to model training and implementation—was planned realistically. The supervisor approved the final proposal, marking a crucial step in officially moving forward with development.
10/02/2025
Meeting #3
Discussed ethics. Since the project involves recording and analyzing video footage, this meeting was centered around ethical considerations, particularly data privacy and informed consent. The supervisor emphasized the importance of ensuring that any recorded clips used for testing and training are either self-generated or acquired with proper consent. Guidelines were discussed regarding the responsible handling of video data, including anonymization practices and secure storage methods to prevent unauthorized access. Additionally, potential concerns about bias in dataset collection were brought up, and steps were outlined to ensure that diverse bowling styles, speeds, and environmental conditions are accounted for in the dataset.
17/02/2025
Meeting #4
Reviewed the literature review. This meeting was dedicated to evaluating the quality and relevance of existing research included in the literature review. The supervisor stressed the importance of analyzing multiple approaches to ball tracking in sports and not just limiting the research to cricket-specific studies. There was also feedback on improving the structure of the review, ensuring a more critical analysis of different tracking techniques, including color filtering, object detection models, and motion tracking. The supervisor recommended incorporating additional studies on how AI-based sports analytics systems have evolved, particularly in tennis and baseball, to gain insights that could be applied to this project.
24/02/2025
Meeting #5
Discussed the software methodology chapter and received approval. The session focused on breaking down the methodology in a clear and structured manner, covering key components such as data collection, preprocessing, model selection, and evaluation criteria. The supervisor provided feedback on improving the explanation of how OpenCV will be used for ball detection, suggesting that a comparison of different detection techniques should be included. We also discussed the iterative approach that would be taken during implementation, ensuring continuous refinement based on test results. After reviewing minor improvements, the methodology chapter was approved, allowing for a smooth transition into the development phase.
03/03/2025
Meeting #6
Discussed UML diagrams and Design Chapter. This meeting revolved around ensuring that the UML diagrams accurately represented the system's architecture and flow of data. The supervisor provided feedback on refining the use case diagrams to better showcase user interactions within the app. We also reviewed the ER diagram, ensuring that the Firebase database structure is well-optimized for handling user data, video clips, and analysis results efficiently. The supervisor highlighted areas where clarity was needed in the design chapter, such as detailing how different components (front-end, database, and AI model) will communicate. This discussion was crucial in finalizing the structural aspects before moving forward with implementation.
10/03/2025
Meeting #7
I met with my supervisor to present the outcomes of the model’s initial training phases. We talked about how I had been running multiple training sessions to get a more refined and consistent output. I showed some early results where the model was able to detect the cricket ball in varying lighting conditions and angles. She pointed out that while progress was clear, it was also important to document changes between each version of training — like what hyperparameters were tweaked or if any image augmentations were added. She also recommended setting up a basic log system to keep track of what version gave the best accuracy and why.
17/03/2025
Meeting #8
During this session, I gave her an update on how I would started building the app itself. I shared my progress using Android Studio and how I was using Visual Studio Code to handle the Firebase connections and real-time updates. We went through how the app retrieves model predictions and displays the ball tracking results. She appreciated the way the app was coming together and reminded me to keep testing Firebase interactions frequently to avoid any surprises later on. She also asked me to check edge cases, like when there's no internet or when data retrieval is delayed, and think of user-friendly ways to handle those moments.
24/03/2025
Meeting #9
​I showed the latest updates to the app and explained how I had fine-tuned the user interface. I’d spent time organizing the layout, testing on different emulators through Android Studio, and ensuring Firebase was reading and writing data accurately. My supervisor suggested implementing progress indicators while the app communicates with Firebase, so users know when the app is processing. She also advised me to clean up unused files and code snippets, especially those that might confuse or slow down performance. I took note of that and decided to spend some time optimizing everything before moving on to the final version.
30/03/2025
Meeting #10
​At this point, the core functions of the app were completed. We discussed final testing and polishing the app’s design. I showed how the app works from the user’s perspective from launching it, to tracking a ball, to displaying the speed and angle data. She was happy with how everything flowed but gave some feedback on layout alignment and font sizes on smaller screens. I told her I’d go through more UX testing and maybe ask a few friends to try the app and share how intuitive it felt. We also touched on the importance of including a help section or tooltips for first-time users, which I plan to implement before the final version.
06/06/2025
Meeting #11
With the app near completion, we shifted focus to the final presentation and report. I shared a draft of my slide deck, and she helped me arrange it in a way that better tells the story of the project from the initial idea, through training, all the way to the working app. She emphasized the importance of showing real examples during the demo instead of just screenshots. We also looked at my written report. She pointed out areas that needed a clearer explanation, like how I used Roboflow to prepare the dataset and what specific challenges I faced during training. I went back and expanded on those parts to make sure everything felt complete and professional.
10/04/2025
Meeting #12
In our last meeting, I did a full walkthrough of the app in real-time. I showed how a user can open the app, choose a video or initiate live tracking, and how the app fetches and displays data using Firebase. She was really pleased with the progress and how every feature tied back to the original goal of real-time ball tracking. We also talked about deployment, she suggested documenting everything from model training to app building clearly, in case I ever want to improve it later or share it with others. I wrapped up by sharing what I learned throughout the journey, and she congratulated me for managing the project well and staying consistent with updates and improvements.
