In many articles (for example, in the YOLO paper, this paper or this one), I see the term "unified" being used. I was wondering what the meaning of "unified" in this case is.
$\begingroup$
$\endgroup$
2
-
$\begingroup$ One example where you see the evolution from "multiple components" to a more "unified neural network" is the family of R-CNN models. If you look at the initial R-CNN, it's composed of more than one component. The successive R-CNNs, such as Fast R-CNN, Faster R-CNN, and Mask R-CNN, if I remember correctly (and maybe not in all cases), try to unify some components to get to a single neural network that is able to do all the intermediate tasks (that should be the Mask R-CNN). I could be wrong because it's been a while since I had to deal with them, but it may be worth reading those papers. $\endgroup$– nbroDec 27, 2020 at 22:54
-
$\begingroup$ @nbro Thank you for the opposite example. $\endgroup$– ReactionicDec 27, 2020 at 23:05
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
A unified neural network model consists of one neural network as opposed to other models that rely on two or more neural networks.
For example, from page two of the YOLO paper:
2. Unified Detection
We unify the separate components of object detection into a single neural network. Our network uses features from the entire image to predict each bounding box. It also predicts all bounding boxes across all classes for an image simultaneously. This means our network reasons globally about the full image and all the objects in the image. The YOLO design enables end-to-end training and realtime speeds while maintaining high average precision.
In the paper by Xu and Wang, they add a branch to handle tracking to the Faster R-CNN architecture, which was designed for object detection. This is a 'unification' of two models.
In the paper by Ebrahimi et al., on pages 42-43, you can see they are fusing multiple neural networks into one unified model:
We propose a unified user geolocation method that relies on a fusion of neural networks, incorporating different types of available information: tweet message, users' social relationships, and metadata fields embedded in tweets and profiles.
