Name
Abstract | ||
|---|---|---|
Given the overwhelming performance of the Convolutional Neural Network (CNN) in the computer vision and machine learning community, this paper aims at investigating the effective transfer of the CNN descriptors in generic and fine-grained classification at a large scale. Our contribution consists in providing some simple yet effective methods in constructing a competitive baseline recognition system. Comprehensively, we study two facts in CNN transfer. (1) We demonstrate the advantage of using images with a properly large size as input to CNN instead of the conventionally resized one. (2) We benchmark the performance of different CNN layers improved by average/max pooling on the feature maps. Our evaluation and observation confirm that the Conv5 descriptor yields very competitive accuracy under such a pooling strategy. Following these good practices, we are capable of producing improved performance on seven image classification benchmarks. |
Year | DOI | Venue |
|---|---|---|
2018 | 10.1007/978-3-030-04503-6_3 | PAKDD |
Field | DocType | Citations |
Recognition system,Convolutional neural network,Computer science,Pooling,Artificial intelligence,Contextual image classification,Transferability,Machine learning | Conference | 0 |
PageRank | References | Authors |
0.34 | 18 | 6 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Liang Zheng | 1 | 1683 | 63.68 |
| Yali Zhao | 2 | 66 | 3.81 |
| Shengjin Wang | 3 | 1450 | 78.26 |
| Jingdong Wang | 4 | 4198 | 156.76 |
| Yi Yang | 5 | 251 | 31.00 |
| Qi Tian | 6 | 6443 | 331.75 |