U-Net with VGG19

U-Net with VGG19 is TT-Metal’s label for an image-segmentation configuration that pairs two separately published architectures. U-Net contributes the general contracting-and-expanding segmentation pattern, while VGG19 names Oxford’s 19-weight-layer convolutional network. The foundational papers do not define this particular pairing, so the safe memory hook is simply ‘U-Net/VGG19 segmentation,’ without assuming undocumented implementation details. Its output is a pixel-level segmentation rather than one class for the whole image. TT-Metal lists 256x256 inputs on n150 Wormhole at batch 1, n300 Wormhole at DP=2 and batch 2, and p150 Blackhole at batch 1. The three rows collapse into this one named configuration.