Image classification is the task of assigning a label from a fixed set of categories to an input image. For example, given an image, decide if it’s a cat, dog, airplane, etc. Datasets like ImageNet, CIFAR-10, MNIST are classic examples for image classification tasks. Modern image classification is dominated by deep learning models, particularly convolutional neural networks (CNNs), which can automatically learn feature hierarchies from pixels to edges to object parts to complete objects. The model is trained on labeled images, adjusting weights to minimize a loss (like cross-entropy) between predicted probabilities and true class labels. At inference, the model outputs a probability distribution over classes, and typically the highest probability is taken as the predicted label. Advances like AlexNet, VGG, ResNet, and EfficientNet have pushed image classification accuracy to very high levels, even surpassing human performance on some benchmarks.