A new approach to descriptors generation for image retrieval by analyzing activations of deep neural network layers
This work addresses image retrieval for computer vision applications, offering an incremental improvement over existing neural code methods.
The paper tackles the problem of constructing effective descriptors for content-based image retrieval by extending neural codes to include convolutional layer activations, proposing an algorithm to extract significant neuron activations. Experimental results on the IMAGENET1M dataset with VGG16 show that the descriptors improve semantic matching and similarity in secondary image characteristics.
In this paper, we consider the problem of descriptors construction for the task of content-based image retrieval using deep neural networks. The idea of neural codes, based on fully connected layers activations, is extended by incorporating the information contained in convolutional layers. It is known that the total number of neurons in the convolutional part of the network is large and the majority of them have little influence on the final classification decision. Therefore, in the paper we propose a novel algorithm that allows us to extract the most significant neuron activations and utilize this information to construct effective descriptors. The descriptors consisting of values taken from both the fully connected and convolutional layers perfectly represent the whole image content. The images retrieved using these descriptors match semantically very well to the query image, and also they are similar in other secondary image characteristics, like background, textures or color distribution. These features of the proposed descriptors are verified experimentally based on the IMAGENET1M dataset using the VGG16 neural network.