Name
Abstract | ||
|---|---|---|
High-quality retrieval techniques can effectively retrieve target images from millions of images, and some classic techniques are widely used in different fields. As a classic image retrieval technique, deep learning shows remarkable advantages in significantly improving retrieval results. However, high-quality retrieval results highly depend on sufficient learning instances. When no sufficient learning instances exist to support learning model construction, then retrieval quality reduces remarkably. In most cases, sufficient learning instances lead to wasting of significant computing and human resources. Aiming at the aforementioned problem, we proposed a weighted-learning-instance-based retrieval model requiring instance distance calculation. Concretely, reference learning instance optimization, instance distance calculation, and innovative cost function construction are combined which could directly contribute to build up the previous model. Firstly, high-quality reference learning instances could be selected by learning instance optimization model. Then, combined with weights of learning instances calculated by instance distance, the innovative cost function could be constructed which could make full use of learning instances under various circumstances. More importantly, this model can significantly reduce the number of learning instances through instance optimization and weight definition while maintaining high level of retrieval quality. Adequate experimental results based on a large database show robustness and effectiveness of our model. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1007/s00138-018-0988-x | Mach. Vis. Appl. |
Keywords | Field | DocType |
Weighted learning instance, Instance distance, Sparse coding, Training function | Pattern recognition,Computer science,Neural coding,Image retrieval,Robustness (computer science),Artificial intelligence,Deep learning,Machine learning | Journal |
Volume | Issue | ISSN |
30 | 1 | 1432-1769 |
Citations | PageRank | References |
0 | 0.34 | 48 |
Authors | ||
7 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Hao Wu | 1 | 143 | 18.69 |
| Yueli Li | 2 | 22 | 2.45 |
| Jie Xiong | 3 | 0 | 0.68 |
| Xiaohan Bi | 4 | 0 | 0.68 |
| Linna Zhang | 5 | 16 | 4.33 |
| Rongfang Bie | 6 | 547 | 68.23 |
| Junqi Guo | 7 | 61 | 15.07 |