The Clara AGX Dermatology Application is a reference application designed for the Clara AGX platform. It leverages computer vision models to detect and classify moles from a webcam feed. The application consists of two primary models: a YoloV4 object detection model for identifying moles and an EfficientNet B0 classification model trained on the HAM10000 dataset to classify moles as Melanoma, Benign, Nevus, or Unknown.

Features

• YoloV4 Object Detection: Detects moles from a webcam feed using the YoloV4 model.
• EfficientNet B0 Classification: Classifies detected moles into categories such as Melanoma, Benign, Nevus, or Unknown.
• HAM10000 Dataset: Utilizes the HAM10000 dataset for training the classification model.
• Jetson Xavier AGX Compatibility: Compatible with Jetson Xavier AGX devices or the Clara AGX Developer Kit running Jetpack 4.5 or later.
• Docker Deployment: Easily deployable using Docker commands.

Benefits

• Accurate Detection and Classification: Provides reliable detection and classification of moles, aiding in dermatological assessments.
• Real-Time Analysis: Enables real-time analysis of webcam feeds for immediate results.
• Ease of Deployment: Simple deployment process using Docker, making it accessible for various users.
• Advanced Training Data: Leverages a well-curated dataset (HAM10000) for accurate classification results.
• Integration with Clara AGX: Seamlessly integrates with the Clara AGX platform, enhancing its capabilities.