Offline AI Disease Detector

I want to build an open-source, end-to-end computer-vision system that can spot diseases across three domains—human healthcare, crop agriculture, and veterinary practice—without ever calling external APIs. Scope • Training: You’ll design and train deep-learning models to recognise Cancer in medical scans, Plant Fungal Infections in leaf images, and Animal Parasitic Infections in animal skin images. Feel free to choose PyTorch, TensorFlow or another modern framework, but the workflow must run entirely on local hardware after setup. • Data: Combine public datasets and clear instructions for adding private data so future users can extend or fine-tune. • Inference: Produce a lightweight CLI (a small GUI is a bonus) that accepts single images or folders, returns class labels and confidence, and saves annotated output images—all offline. • Source code: Clean, well-commented, published under an OSI-approved licence. • Documentation: Step-by-step guides covering environment setup, data preparation, training, model evaluation, and deployment on Windows, macOS, and Linux. Acceptance A fresh machine with no internet access must: 1. Reproduce the training run from the provided scripts. 2. Classify a test set from each domain with accuracy comparable to your validation results. 3. Launch the inference tool in under one minute and predict locally. If this matches your skill set in deep learning, computer vision, and cross-domain data handling, let’s get started.

Project ID: 40595348