Learning Deep Features for Discriminative Localization
share
In this work, we revisit the global average pooling layer proposed in [13], and shed light on how it explicitly enables the convolutional neural network to have remarkable localization ability despite being trained on image-level labels. While this technique was previously proposed as a means for regularizing training, we find that it actually builds a generic localizable deep representation that can be applied to a variety of tasks. Despite the apparent simplicity of global average pooling, we are able to achieve 37.1 top-5 error for object localization on ILSVRC 2014, which is remarkably close to the 34.2 demonstrate that our network is able to localize the discriminative image regions on a variety of tasks despite not being trained for them
READ FULL TEXT