PCANet is a simple deep learning network for image classification using cascaded principal component analysis (PCA), binary hashing, and block-wise histograms. It outperforms state-of-the-art features in tasks like face recognition, handwritten digit recognition, and texture classification. Variations like RandNet (random filters) and LDANet (filters learned from LDA) are also studied. PCANet achieves high accuracy on datasets such as LFW, Extended Yale B, AR, FERET, and MNIST. It is efficient, easy to train, and robust to variations in lighting, occlusion, and deformation. PCANet's simplicity and effectiveness make it a competitive baseline for image classification tasks.PCANet is a simple deep learning network for image classification using cascaded principal component analysis (PCA), binary hashing, and block-wise histograms. It outperforms state-of-the-art features in tasks like face recognition, handwritten digit recognition, and texture classification. Variations like RandNet (random filters) and LDANet (filters learned from LDA) are also studied. PCANet achieves high accuracy on datasets such as LFW, Extended Yale B, AR, FERET, and MNIST. It is efficient, easy to train, and robust to variations in lighting, occlusion, and deformation. PCANet's simplicity and effectiveness make it a competitive baseline for image classification tasks.