# ASIC-RAG: Hardware-Accelerated Medical Anomaly Detection and Cryptographic Data Sovereignty This repository contains the full codebase for the **ASIC-RAG-CHIMERA** framework, focusing on medical anomaly detection (Pneumonia) using hybrid ASIC-CNN architectures. ## Repository Structure - `Firmware_ASIC_LV06/`: Drivers and firmware for the Lucky Miner LV06 hardware. - `V4/`: The latest multi-scale hybrid model implementation, including trainer, models, and evaluation scripts. - `ASIC-RAG-CHIMERA_...pdf`: Academic whitepaper detailing the framework. - `ai_studio_code.html`: Interactive research report. ## Dataset Information To keep the repository lightweight and comply with GitHub file size limits, the radiographs dataset is **not included** in this repository. ### Download Dataset The model is trained on the **Pneumonia X-Ray Dataset**. You can download it directly from Hugging Face: **Source:** [mmenendezg/pneumonia_x_ray](https://huggingface.co/datasets/mmenendezg/pneumonia_x_ray) After downloading, place the dataset in the `V4/data/` directory with the following structure: ``` V4/data/ ├── train/ ├── val/ └── test/ ``` ## Hardware Requirements This system is designed to interface with **Lucky Miner LV06 (BM1366 ASIC)** for hardware-accelerated attention maps and cryptographic verification. ## Author Francisco Angulo de Lafuente