View the presentation here
Developed a web-based product, Skin Zen, that accurately detects and classifies the severity of acne using the ResNet-101 architecture. The model was pre-trained with weights from the ImageNet dataset. The project incorporated additional models like eye-cascade and face-landmarks to extract skin patches from training images, enhancing the classification process.
Implemented a recommendation component in Skin Zen using the GPT-3 API. Leveraged the GPT-3 language model to generate human-like text responses. The recommendation system utilized GPT-3 by providing consistent prompts on how the severity detection model works and how to respond.
