On the large-scale transferability of convolutional neural networks
- Publication Type:
- Conference Proceeding
- Citation:
- Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2018, 11154 LNAI pp. 27 - 39
- Issue Date:
- 2018-01-01
Closed Access
| Filename | Description | Size | |||
|---|---|---|---|---|---|
| 2018_Book_TrendsAndApplicationsInKnowled.pdf | Published version | 20.73 MB |
Copyright Clearance Process
- Recently Added
- In Progress
- Closed Access
This item is closed access and not available.
© Springer Nature Switzerland AG 2018. 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.
Please use this identifier to cite or link to this item: