Efficient Deep Learning with Narrow ResNet-18

Built a compact and high-performing image classification model using a width-scaled ResNet-18 architecture on the CIFAR-10 dataset. Optimized for efficiency by reducing model size under 5M parameters and incorporating Cutout augmentation, label smoothing, and cosine annealing learning rate scheduling. Achieved 95.5% test accuracy in 250 epochs while significantly lowering memory and compute requirements, making it suitable for edge deployment.