Object Localization with a Weakly Supervised CapsNet
This addresses object localization for computer vision applications, offering a novel method that is incremental by building on CapsNet with modifications.
The paper tackles the problem of object localization using a weakly supervised CapsNet, achieving state-of-the-art performance by learning and deriving object coordinates from class labels without extra localization modules, as demonstrated on moving MNIST and KTH datasets.
Inspired by CapsNet's routing-by-agreement mechanism with its ability to learn object properties, we explore if those properties in turn can determine new properties of the objects, such as the locations. We then propose a CapsNet architecture with object coordinate atoms and a modified routing-by-agreement algorithm with unevenly distributed initial routing probabilities. The model is based on CapsNet but uses a routing algorithm to find the objects' approximate positions in the image coordinate system. We also discussed how to derive the property of translation through coordinate atoms and we show the importance of sparse representation. We train our model on the single moving MNIST dataset with class labels. Our model can learn and derive the coordinates of the digits better than its convolution counterpart that lacks a routing-by-agreement algorithm, and can also perform well when testing on the multi-digit moving MNIST and KTH datasets. The results show our method reaches the state-of-art performance on object localization without any extra localization techniques and modules as in prior work.