Abstract | ||
---|---|---|
Image search can be tackled using deep features from pre-trained Convolutional Neural Networks (CNN). The feature map from the last convolutional layer of a CNN encodes descriptive information from which a discriminative global descriptor can be obtained. We propose a new representation of co-occurrences from deep convolutional features to extract additional relevant information from this last convolutional layer. Combining this co-occurrence map with the feature map, we achieve an improved image representation. We present two different methods to get the co-occurrence representation, the first one based on direct aggregation of activations, and the second one, based on a trainable co-occurrence representation. The image descriptors derived from our methodology improve the performance in very well-known image retrieval datasets as we prove in the experiments. |
Year | DOI | Venue |
---|---|---|
2020 | 10.1016/j.imavis.2020.103909 | Image and Vision Computing |
Keywords | DocType | Volume |
Co-occurrence,Image retrieval,Feature aggregation,Pooling | Journal | 97 |
ISSN | Citations | PageRank |
0262-8856 | 0 | 0.34 |
References | Authors | |
0 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Forcen J. I. | 1 | 0 | 0.34 |
Miguel Pagola | 2 | 979 | 45.68 |
Barrenechea Edurne | 3 | 0 | 0.34 |
Humberto Bustince | 4 | 1938 | 134.10 |