Post-Training INT8 Quantization of MobileNetV3 for Efficient Chest X-Ray Classification on Edge Devices

This research investigates post-training INT8 quantization of MobileNetV3 for efficient chest X-ray classification on resource-constrained edge devices. The study provides a comparative evaluation between the original FP32 and quantized INT8 models in terms of classification performance, model size, and inference latency. The proposed approach achieves substantial model compression while maintaining competitive diagnostic performance, demonstrating the potential of post-training quantization for developing lightweight and practical AI-based medical image classification systems for edge deployment.