I need a skilled programmer to build and train an AI model specifically for detecting cracks on aircraft metal panels using high-resolution still images. You will also develop a Python API to connect the trained model to my existing web app for automated visual inspection. Key Requirements: - Build and train an AI model using existing labeled images. - Detect cracks in high-resolution still images. - Develop a Python API integrating with my web app. Ideal Skills and Experience: - Expertise in AI/ML, particularly in image recognition. - Strong experience with Python and API development. - Familiarity with automated visual inspection systems. - Prior experience in the aerospace industry is a plus. AI-Based Crack Detection Using Camera and Image Processing This project aims to develop an AI-based visual inspection system that uses camera images to automatically detect cracks and surface defects on aircraft metal panels. A deep learning model is trained on labeled images of surface cracks and is used to analyse new images captured by a camera or uploaded by the user. The system outputs the detection result in the form of: • the location of detected cracks on the image, and • a final inspection decision for each image (crack detected or no crack detected). The purpose of the system is to support aircraft maintenance engineers by improving the speed and reliability of visual inspection and reducing human error. The main focus is on building and training a new AI model for crack and surface defect detection on metal panels. • The developer should prepare and fine-tune a suitable deep learning model or similar (CNN-based detector) using existing labeled datasets. • The final deliverable should include the trained model, training code, and a simple script to run inference on new images. • A short explanation of the model, training process, and evaluation results is required for academic use. After analysis, the system must provide a clear result for each image (Crack detected / No crack detected), in addition to the detection boxes and confidence scores.