OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks

We present an integrated framework for using Convolutional Networks for, localization and detection. We show how a multiscale and window approach can be efficiently implemented within a ConvNet. We introduce a novel deep learning approach to localization by learning to object boundaries. Bounding boxes are then accumulated rather than in order to increase detection confidence. We show that different can be learned simultaneously using a single shared network. This framework is the winner of the localization task of the ImageNet Scale Visual Recognition Challenge 2013 (ILSVRC2013) and obtained very results for the detection and classifications tasks. In-competition work, we establish a new state of the art for the detection. Finally, we release a feature extractor from our best model called.