MuCaLe-Net: Multi Categorical-Level Networks to Generate More Discriminating Features

Abstract

In a transfer-learning scheme, the intermediate layers of a pre-trained CNN are employed as universal image representation to tackle many visual classification problems. The current trend to generate such representation is to learn a CNN on a large set of images labeled among the most specific categories. Such processes ignore potential relations between categories, as well as the categorical-levels used by humans to classify. In this paper, we propose Multi Categorical-Level Networks (MuCaLe-Net) that include human-categorization knowledge into the CNN learning process. A MuCaLe-Net separates generic categories from each other while it independently distinguishes specific ones. It thereby generates different features in the intermediate layers that are complementary when combined together. Advantageously, our method does not require additive data nor annotation to train the network. The extensive experiments over four publicly available benchmarks of image classification exhibit state-of-the-art performances.

DeeSIL: Deep-Shallow Incremental Learning

“…iCaRL halves this score while our best configurations with DF E DF E DF E based on 100 and 1000 classes lose only 22 and 12 points respectively. The gap could probably be further reduced if the feature extractors were more universal [10,11]. This could, for instance, be achieved if DeeSIL's initial training would be done with an even larger number of classes.…”

“…One thousand classes are selected to form a diversified subset of ImageNet and thus increase universality (i.e. optimize their transferability toward new tasks) [10,11]. -F L1000 -train with a more challenging dataset which is obtained from weakly annotated Flickr group data and is visually more distant from the test set.…”

Joint Learning of Pre-Trained and Random Units for Domain Adaptation in Part-of-Speech Tagging

“…Bias when fine-tuning pre-trained units Here our goal is to highlight that as in (Zhou et al, 2018a), pre-trained units can be biased in the standard fine-tuning scheme. To do so, we follow (Tamaazousti et al, 2017) and analyse the units of Φ (biLSTM layer) before and after finetuning. Specifically, we compute the Pearson's correlation between all the units of the layer before and after fine-tuning.…”

Learning Finer-class Networks for Universal Representations

“…Although this similarity is not easy to formalize, one has the intuition that the closer the both tasks the better the representation will be adapted to the target-task. This consideration leads to several methods that tend to obtain more universal representations [2,9,19,29,35,37], that is to say that are more adapted to a large set of diverse target-tasks, in a transfer-learning scenario.…”

“…All these approaches vary the problem by creating new categories having an existing label. However, most of them studied the effect of adding categories extracted from ImageNet, either generic categories [20,23,35,37] or specific ones [1,2,29,46], that are at the bottom of a hierarchy such as ImageNet, except [18,38,39] that use web annotations with noisy labels. In general, the usage of specific categories tends to provide better performances than generic ones [2,29], although combining them can significantly boost the universalizing capacity of the CNN [35,37].…”