Paper List
- A Theoretically Sound Upper Bound on the Triplet Loss for Improving the Efficiency of Deep Distance Metric Learning
- End-to-End Supervised Product Quantization for Image Search and Retrieval
- Ranked List Loss for Deep Metric Learning
- On Learning Density Aware Embeddings
- Stochastic Class-based Hard Example Mining for Deep Metric Learning
- Multi-Similarity Loss with General Pair Weighting for Deep Metric Learning
- Deep Metric Learning to Rank
- Learning Metrics from Teachers: Compact Networks for Image Embedding
- Deep Embedding Learning with Discriminative Sampling Policy
- Divide and Conquer the Embedding Space for Metric Learning
- Unsupervised Embedding Learning via Invariant and Spreading Instance Feature
- Signal-to-Noise Ratio: A Robust Distance Metric for Deep Metric Learning
- Deep Asymmetric Metric Learning via Rich Relationship Mining
- Hardness-Aware Deep Metric Learning
数据集及评价指标：
CUB-200-2011
| Method | R@1 | R@2 | R@4 | R@8 |
| 1. Discriminative | 51.43 | 64.23 | 74.31 | 82.83 |
| 3.RLL-(L,M,H) | 61.3 | 72.7 | 82.7 | 89.4 |
| 5.SCHE | 66.2 | 76.3 | 84.1 | 90.1 |
| 6.MS | 65.7 | 77.0 | 86.3 | 91.2 |
| 9. DE-DSP (N-pair) | 53.6 | 65.5 | 76.9 | - |
| 10. DCES | 65.9 | 76.6 | 84.4 | 90.6 |
| 12. DSML | 51.6 | 54.9 | - | - |
| 13. RRM | 55.1 | 66.5 | 76.8 | 85.3 |
| 14. HDML | 53.7 | 65.7 | 76.7 | 85.7 |
CAR196
| Method | R@1 | R@2 | R@4 | R@8 |
| 1. Discriminative | 68.31 | 78.21 | 85.22 | 91.18 |
| 3.RLL-(L,M,H) | 82.1 | 89.3 | 93.7 | 96.7 |
| 5.SCHE | 91.7 | 95.3 | 97.3 | 98.4 |
| 6.MS | 84.1 | 90.4 | 94.0 | 96.5 |
| 9. DE-DSP (N-pair) | 72.9 | 81.6 | 88.8 | - |
| 10. DCES | 84.6 | 90.7 | 94.1 | 96.5 |
| 12. DSML | 49.1 | 52.4 | - | - |
| 13. RRM | 73.5 | 82.6 | 89.1 | 93.5 |
| 14. HDML | 79.1 | 87.1 | 92.1 | 95.5 |
SOP
| Method | R@1 | R@10 | R@100 |
| 3.RLL-(L,M,H) | 79.8 | 91.3 | 96.3 |
| 5.SCHE | 77.6 | 89.1 | 94.7 |
| 6.MS | 78.2 | 90.5 | 96.0 |
| 7.FastAP | 75.8 | 89.1 | 95.4 |
| 9. DE-DSP (N-pair) | 68.9 | 84.0 | 92.6 |
| 10. DCES | 75.9 | 88.4 | 94.9 |
| 13. RRM | 69.7 | 85.2 | 93.2 |
| 14. HDML | 68.7 | 83.2 | 92.4 |
In-shop
| Method | R@1 | R@10 | R@20 | R@30 |
| 5.SCHE | 91.9 | 98.0 | 98.7 | 99.0 |
| 6.MS | 89.7 | 97.9 | 98.5 | 98.8 |
| 7.FastAP | 90.9 | 97.7 | 98.5 | 98.8 |
| 9. DE-DSP (N-pair) | 78.6 | 93.8 | 95.5 | 96.2 |
| 10. DCES | 85.7 | 95.5 | 96.9 | 97.5 |
以上就是本文的全部内容，希望对大家的学习有所帮助，也希望大家多多支持 码农网
猜你喜欢:
- 【docker 笔记】dockerize 相关整理
- JVM的相关知识整理和学习
- 原 荐 vue打包相关细节整理
- 阿里、字节：一套高效的 iOS 面试题之我整理的答案之 runtime 相关问题（一）
- 照片整理系列二 —— 照片整理及归档的辛酸历程
- 我自己整理的码农周刊一周分享整理
本站部分资源来源于网络，本站转载出于传递更多信息之目的，版权归原作者或者来源机构所有，如转载稿涉及版权问题，请联系我们。
Java 8实战
厄马(Raoul-Gabriel Urma)、弗斯科(Mario Fusco)、米克罗夫特(Alan Mycroft) / 陆明刚、劳佳 / 人民邮电出版社 / 2016-4-1 / CNY 79.00
本书全面介绍了Java 8 这个里程碑版本的新特性，包括Lambdas、流和函数式编程。有了函数式的编程特性，可以让代码更简洁，同时也能自动化地利用多核硬件。全书分四个部分：基础知识、函数式数据处理、高效Java 8 编程和超越Java 8，清晰明了地向读者展现了一幅Java 与时俱进的现代化画卷。一起来看看 《Java 8实战》 这本书的介绍吧!
