Hello everyone, here is [Come to a Scallion Cake], this time I brought a technical sharing of quick introduction to image retrieval , welcome to pay attention, and share with you~
Interested students, after reading this article, you can also read my article Review interpretation, more comprehensive, image retrieval｜Classic paper reading｜Quick start｜Review learning
I. Overview
1. Image retrieval is to search for images by image
2. Mainly divided into two categories.
The first one is to pursue the so-called similarity, that is, to find similar or identical ones from the gallery. Common face recognition, pedestrian re-identification, commodity retrieval, and sketch retrieval all belong to this category. The current deep learning method is more More is to pursue how to obtain better features. In these fields, the methods are actually universal .
The second is a match problem that needs to be done in the traditional field, that is, if a certain building appears in one picture, if it only appears half of it in another picture, or appears in a certain corner, can it be found again? . If the cnn feature is used directly, it is very likely to find buildings with similar outlines, not the same buildings. In this case, it is generally cut into patches and then extract features separately. Of course, after the faster rcnn comes out later, it can be directly divided on the final feautre map . It is also possible to perform vlad or fv encoding and retrieval on the features of the feature map. The titles in general papers are CBIR, instance retrieval, etc.
3. Image retrieval includes content-based retrieval (nlp) and image-based retrieval (cv). We are talking about cv
4. Image retrieval has not made much progress in recent years, but people have been doing it.
2. Image retrieval sota model
From paperswithcode.com:
3. Related Blogs
So far (2020), what excellent research work has been done in the field of image retrieval (Image Retrieval) over the years? - WALL-E's answer- Zhihu
Also, there is a good review: SIFT Meets CNN: A Decade Survey of Instance Retrieval – published in 2017
Code words are not easy, if you have seen this, why don’t you give a thumbs up~
Here is [ Come on a Scallion Cake ], your likes + favorites + attention are the biggest motivation for me to persevere~
