CNNs & Computer Vision

    AI · Advanced

    Definition

    Convolutional Neural Networks slide small filters over images to detect edges, shapes and objects.

    CNNs drive face unlock, medical imaging and self-driving perception.

    Diagram

    Image→ Conv→ Pool→ Dense → 'cat'

    Flowchart

    Input imageConvolution filtersReLUPoolingFlattenDense layer → class

    Example: Simple 1D convolution

    import numpy as np
print(np.convolve([1,2,3,4], [1,-1], 'valid'))

    Output: [1 1 1]

    Key points

    • Weight sharing reduces parameters
    • Pooling adds translation tolerance
    • Transfer learning reuses pretrained models

    Interview questions

    What does pooling do?

    Downsamples feature maps, keeping the strongest signals.

    More AI topics

    What is Artificial IntelligenceNeural NetworksGradient DescentNLP, Transformers & LLMs