CNN

Motivation Ensure shift-invariance The model should produce the same output regardless of the position of the considering object Overview Multilayer Neural Network: Nonlinear Classifier Consider Context (Receptive Field)

Architecture Overview All CNN models follow a similar architecture Input Convolutional layer (Cons-layer) + ReLU Pooling layer (Pool-layer) Fully Connected layer (FC-layer) Output Input The input layer represents the input image into the CNN.

Intuition A CNN model can be thought as a combination of two components: feature extraction part The convolution + pooling layers perform feature extraction. For example given an image, the convolution layer detects features such as two eyes, long ears, four legs, a short tail and so on.

LeNet (1998) Image followed by multiple convolutional / pooling layers Build up hierarchical filter structures Subsampling / pooling increases robustness Fully connected layers towards the end

Computer Vision (CV) Tasks Classification Classification + Localization Object Detection Instance Segmentation Object Localization: Coordinate prediction Sliding Window Object Localization Classification & Localization Detection Sliding Window + Classification:

Tutorials How do Convolutional Neural Networks work? An Intuitive Explanation of Convolutional Neural Networks Visualization CNN Explainer MINST playground Plotting NN-SVG Papers Overview The 9 Deep Learning Papers You Need To Know About (Understanding CNNs Part 3)