U-Net Segmentation
A neural network architecture that outputs pixel-wise classification maps.
U-Net is an encoder-decoder architecture designed to output a segmentation mask, mapping every pixel of an input image to a specific class.
The left side of the network (encoder) gradually reduces the spatial dimensions of the image while increasing the number of feature channels, capturing the "what" (context).
The right side (decoder) progressively restores the spatial dimensions, reconstructing the "where" (localization) to output a full-resolution mask.
To recover fine-grained spatial details lost during downsampling, U-Net uses skip connections. These transfer high-resolution features directly from the encoder to the decoder, merging context with precise localization.