When Naïve Bayes Nearest Neighbours Meet Convolutional Neural Networks
This work revives NBNN classifiers for visual recognition by making them compatible with modern CNN architectures, addressing a specific bottleneck for researchers in computer vision.
The paper tackled the problem of integrating Naive Bayes Nearest Neighbour (NBNN) classifiers with Convolutional Neural Networks (CNNs) by proposing a framework that enables NBNNs to use CNN activations, be trained end-to-end, and scale to big data, resulting in improved performance on standard scene and domain adaptation databases.
Since Convolutional Neural Networks (CNNs) have become the leading learning paradigm in visual recognition, Naive Bayes Nearest Neighbour (NBNN)-based classifiers have lost momentum in the community. This is because (1) such algorithms cannot use CNN activations as input features; (2) they cannot be used as final layer of CNN architectures for end-to-end training , and (3) they are generally not scalable and hence cannot handle big data. This paper proposes a framework that addresses all these issues, thus bringing back NBNNs on the map. We solve the first by extracting CNN activations from local patches at multiple scale levels, similarly to [1]. We address simultaneously the second and third by proposing a scalable version of Naive Bayes Non-linear Learning (NBNL, [2]). Results obtained using pre-trained CNNs on standard scene and domain adaptation databases show the strength of our approach, opening a new season for NBNNs.