基于神经网络的图像检索 Neural Codes for Image Retrieval
来源:互联网 发布:淘宝助理 for mac 编辑:程序博客网 时间:2024/05/17 04:26
Neural Codes for Image Retrieval
An example of the retrieval with neural codes. Each row shows the query on the left and the images from the INRIA dataset with the most similar neural codes to the right. The three rows correspond to the three layers in the convolutional neural network.
In this project, we investigate what are the best ways to get global (holistic) image descriptors out of deep neural networks. We aim at relatively low-dimensional descriptors (e.g. 128-256 dimensions to describe an image).
The current “best” way that we have found is summarized in the ICCV 2015 paper. In short, the descriptors are based on the last convolutional layer of a pretrained deep network, sum-pooling, and simple post-processing. We call the resulting descriptor SPoC (for Sum-Pooling of Convolutional features).
In the ICCV paper, we obtained very competitive retrieval accuracy while using a network pretrained on ImageNet and not fine-tuned for buildings/landmarks. However, in our earlier ECCV 2014 paper, we have shown that such fine-tuning is beneficial and have collected a special dataset suitable for such finetuning (the Landmarks dataset). Note that at the moment we recommend SPoC (and similar) descriptors over using outputs of the fully-connected layers (as in the ECCV 2014 paper) when generic image descriptors are required. Fine-tuning on related dataset should be used if maximal performance is desired, as our preliminary experiments suggest that SPoC also benefit from fine-tuning though to a lesser degree than fully-connected features.
Papers:
A. Babenko and V. Lempitsky. Aggregating local deep features for image retrieval. IEEE International Conference on Computer Vision (ICCV), Santiago de Chile, 2015
arXiv:1510.07493
A. Babenko, A. Slesarev, A. Chigorin and V. Lempitsky. Neural Codes for Image Retrieval, European Conference on Computer Vision (ECCV), Zurich, 2014
Paper
Code:
The code and usage example for the SPOC descriptor: https://github.com/arbabenko/Spoc
The Landmarks Dataset:
Tab-separated file
Format of each line: url image_id class_name (in Russian)
Note: Oxford-related classes were manually removed for our experiments in the ECCV 2014 paper.
Related work:
Several other groups have been investigating similar directions recently and in parallel. Some very interesting results have been presented by the KTH Stockholm group:
http://www.csc.kth.se/cvap/cvg/DL/ots/
from: http://sites.skoltech.ru/compvision/projects/neuralcodes/
- 基于神经网络的图像检索 Neural Codes for Image Retrieval
- 基于deep learning的快速图像检索(Deep Learning of Binary Hash Codes for Fast Image Retrieval)
- 基于deep learning的快速图像检索(Deep Learning of Binary Hash Codes for Fast Image Retrieval)
- 深度学习用于基于内容的图像检索 Deep Learning for Content-Based Image Retrieval
- 基于内容的图像检索 Database for Content-Based Image Retrieval
- 基于哈希的图像检索 HABIR(HASHING BASELINE FOR IMAGE RETRIEVAL)
- 图像检索系列一:Deep Learning of Binary Hash Codes for Fast Image Retrieval
- 图像检索系列一:Deep Learning of Binary Hash Codes for Fast Image Retrieval
- 快速图像检索(Deep Learning of Binary Hash Codes for Fast Image Retrieval)
- 图像检索--Deep Supervised Hashing for Fast Image Retrieval
- 基于内容的图像检索技术(Content-Based Image Retrieval,CBIR)
- 基于Bow模型的图像检索 Image Retrieval with Bag of Visual Words
- 一篇论文 综合区域匹配的图像检索 IRM: Integrated Region Matching for Image Retrieval
- 图像检索系统《Deep Learning of Binary Hash Codes for Fast Image Retrieval》
- 大规模图像数据库的搜索-图像检索 Searching in large image databases -- Image retrieval
- 卷及神经网络CNN for image retrieval
- 局部特征用于图像检索 Aggregating local features for Image Retrieval
- Deep Learning of Binary Hash Codes for Fast Image Retrieval
- 百度地图ASP.NET使用例子
- (java)Combination Sum II
- OpenWrt固件编译刷机完全总结
- laravel使用mysql队列
- android应用程序第一次启动时显示引导界面
- 基于神经网络的图像检索 Neural Codes for Image Retrieval
- 如何快速启动mysql等数据库
- 【剑指offer】5.3时间效率与空间效率的平衡——面试题34:丑数
- Spark性能优化——序列化、内存、并行度、数据存储格式、Shuffle
- NumPy的详细教程
- Android 中几个Context的区别,getApplicationContext(),Activity.this,getBaseContext()
- red5 + rtmp篇2
- POJ 1789 Truck History
- UIVIEW 设置圆角 +边框阴影