The last two weeks of research in ISM have been pivotal in shaping the direction of my project. I reviewed two research papers that provided significant insights into both the technical aspects of AI-driven skin cancer detection and the practical considerations of mobile app development. The first paper, “Skin Cancer Recognition Using Unified Deep Convolutional Neural Networks,” introduced me to the YOLO (You Only Look Once) models, which are widely used for real-time object detection. After comparing YOLOv3, YOLOv4, YOLOv5, and YOLOv7, I learned that YOLOv7 was the best-performing model in terms of accuracy and speed, making it an ideal choice for my app. The study also highlighted the importance of data augmentation techniques like flipping, rotating, and adjusting brightness, which I plan to use to improve the robustness of my model. This understanding has helped me rethink how I can optimize the AI for detecting and classifying skin lesions in real-time. The second paper, “Cell Phone Usefulness to Improve Skin Cancer Screening: Preliminary Results and Critical Analysis of Mobile App Development,” gave me a clear idea of how to refine the app’s user experience. The research emphasized the importance of capturing high-quality images from multiple angles, which I will incorporate into the app by guiding users through a photo-taking process that ensures diagnostic accuracy. I also learned that teledermatology, or remote dermatology consultations through smartphones, is highly reliable, particularly in underserved regions where access to specialists is limited. This aligns with my goal of making the app accessible to all, especially in rural areas. One of the key takeaways was the need to ensure high-resolution images for accurate AI analysis, which shifted my focus toward building an image quality assessment tool within the app. The idea is to give users real-time feedback on their photos, prompting them to retake any poor-quality images before they’re submitted for analysis. These two research papers have not only validated many of my initial ideas but also opened up new possibilities for improving my app’s functionality and accessibility. I am excited to continue refining the project based on these insights, and I look forward to implementing YOLOv7 in my model while ensuring the app provides real-time, accurate feedback to users for early skin cancer detection.
