Flexing my AI knowledge in the aesthetic and beauty domain. This web app detects the presence of gynecomastia in men. I have created this web app https://isitgyno.com
What is Gynecomastia?
Gynecomastia is the medical term for enlarged male breast tissue—commonly referred to as "man-boobs". It gives the chest a more feminine appearance and can affect self-esteem. There are four grades depending on how severe the case is. True gynecomastia cannot be removed through exercise or weight loss due to the presence of breast glands in the chest. This condition creates insecurity for some men. The only definitive treatment is surgery—specifically, the removal of the breast gland.
How to use?
Visit https://isitgyno.com, upload a clear front-facing photo, and receive instant AI feedback on whether signs of gynecomastia are detected.
How did I build it?
My core expertise is in web development, particularly with Drupal and backend systems. To bring this idea to life, I had to dive deep into the world of Python and machine learning—specifically convolutional neural networks (CNNs).
It took me about two months to build a working AI model. I leaned heavily on ChatGPT throughout the process, not just for code, but for understanding concepts I wasn’t initially familiar with. There were moments when I had to pause, zoom out, and really understand what was happening under the hood before moving forward.
The tech stack
I’ve built a full-stack gynecomastia detection web app combining modern frontend and AI tooling:
💻 Frontend: SvelteKit
- I use SvelteKit for the web interface. It’s fast, lightweight, and perfect for building reactive, SEO-friendly web applications.
- Styling is handled via Tailwind CSS for responsiveness and consistent UI across devices.
- Users can securely upload a photo and receive a fast AI-driven assessment.
🧠 AI Model: PyTorch CNN Classifier
- The core engine is a convolutional neural network (CNN) model built with PyTorch, trained to detect visual indicators of gynecomastia in chest-area images.
- I’ve used ResNet-50 as the backbone, with a custom classification head fine-tuned on curated datasets.
- Preprocessing is handled using Albumentations, and inference is optimized for performance and minimal false positives.
⚙️ Backend & Deployment
- Image uploads and AI predictions are handled via a Python backend (serverless or containerized depending on environment).
- All uploaded photos are handled securely.
- The model inference can run on a local server or a hosted cloud inference platform like SageMaker or Docker-based containers.
Try it out for yourself!