We're excited to continue advancing healthcare experiences using the latest advances in AI and CV.
As I conclude my internship at Ponce AI, I wanted to share some insights into the fascinating projects I’ve had the privilege to work on, especially the Lip Type Clustering initiative, which was both challenging and immensely rewarding. Here's a glimpse into this innovative endeavor: ? - Data Collection: Embraced the FFHQ dataset along with diverse user-uploaded images to ensure a broad spectrum of lip types. - Pre-Processing: Used Mediapipe for precise lip region detection and cropping, focusing on the area crucial for accurate classification. - Efficient Clustering: Implemented Annoy Index for organizing lip images into clusters based on similarity, which was pivotal for categorizing different lip structures. - Quality Evaluation and Insight: Initially used Dunn Index and Silhouette Score, then transitioned to a more direct measurement of cluster relevance based on centroid distance. - Outcome: Successfully categorized lips into distinct clusters, significantly enhancing the realism of cosmetic simulations. - Operational Workflow: This clustering system can be integrated within the image generation pipeline, allowing for any input image to be automatically assigned to a cluster. This assignment can then inform the subsequent simulation processes, ensuring that each simulation was tailored to the specific characteristics of the input image. Other Noteworthy Projects: - Analyzed input image quality to differentiate between sharp and blurry images, ensuring high-quality inputs for the stable diffusion pipeline. - Developed an AI-powered RAG-based chatbot for medical consultations, improving customer interactions and query handling. I am immensely grateful to sugam arora, Rudolph Layko, and the entire team at Ponce AI for their guidance and mentorship. This experience has been a significant milestone in my professional journey, and I am excited to carry these learnings forward in my career. #DataScience #MachineLearning #ComputerVision #imageprocessing #InternshipExperience #InnovationInTech #GenerativeAI #stablediffusion #largelanguagemodels #chatbot