How does the faster R-CNN model work?
In: Technology
A neural network works by flowing a bunch of inputs (anything, really, but for this stuff it’s each pixel of an image) through nodes to outputs (decisions). The nodes perform some basic math combining inputs. They giggle the weights or coefficients for every node according to some math about how far off the output is until the output is better.
Convolution is a mathmathical term for how two things get combined and affect each other. Pertaining to neural networks, it’s a layer before you get to those raw math nodes to help bots match things. ie, “this is how this image would affect that image”.
“With regions”, means you take the whole image-recognition CNN and test it on a specific rectangle of an image.
“FASTER” just means they’ve got some shortcuts to help pick the region instead of trying random arbitrary rectangles in the image.
So….. Look at an area of a picture. Does it share any features with the thing you’re looking for? Is this what you’re looking for?
Latest Answers