fully convolutional network

A convolutional neural networkconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects. built without fully connected layersfully connected layer, n.: A layer in which every input is connected to every output by its own adjustable weight. It is the most general and most common kind of neural-network layer., from convolutionalconvolution, n.: An operation that slides a small grid of weights across an image or signal, multiplying and adding at each spot, to measure how strongly a pattern appears at every location., poolingpooling, n.: A step in an image network that shrinks its internal picture by summarizing each small patch with one number, usually the largest or the average, keeping what matters and discarding exact positions. and upsampling layers alone, so that for an input of whatever size it returns a map of predictions, one per location, rather than a single label. Rewriting a classifier's fully connected layers as equivalent convolutions lets it be applied densely across a whole image in one pass. For semantic segmentationsemantic segmentation, n.: Coloring every pixel of an image by what it shows, such as road, sky, person or car, without telling one person or car apart from another., coarse deep feature mapsfeature map, n.: The grid of numbers produced when one learned pattern detector is scanned across an image, showing where, and how strongly, that pattern appears. are upsampled, by transposed convolutionstransposed convolution, n.: An operation that roughly reverses an image network's shrinking steps, enlarging a small grid of features into a bigger one, as when a model paints a full-size image. or interpolation, and fused with finer maps from earlier layers through skip connections, yielding a class prediction for every pixel. The design underlies most later segmentation and dense-prediction networks, including U-NetU-Net, prop. n.: An image network shaped like the letter U: one half shrinks the picture to understand it, the other enlarges it back, with shortcuts between the halves so fine detail is not lost., and the convolutional backbones of many object detectors.

Because the model is fully convolutional, we can run it on the full 4K frame without resizing.