Series of RCNN-优快云博客

本文介绍了一种基于图论的图像分割方法，该方法通过构建图像的图表示来进行高效的分割处理。此外，还讨论了一个针对目标检测和分割的系统R-CNN，它利用卷积神经网络从候选区域提取特征，并通过线性SVM进行分类。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

Title

Efficient Graph-Based Image Segmentation

Link

International Journal of Computer Vision 59(2), 167-181,2004

Abstract

this research belong to image segmentation; its principle is a graph-based representation of the image.

Graph
- G = (V,E). V and E represent vertices and edges respectively
- internal difference.laegest weight in the minimum spanning tree of the component MST(C,E): Int(C) = max w(e)

Reference Bolg

https://blog.youkuaiyun.com/guoyunfei20/article/details/78727972

Title

Rich feature hierarchies for Accurate Object Detection and Segmentation

Link

arXiv:1311.2524v5

Abstract

this paper aim to object detection. it has two key insights: 1) they apply convolutional neural networks to bottom-up region proposals in order to localize and segment objects. 2) fine-tuning will make a significant performance boost. They call their methods R-CNN

their system has three modules: the first generate category-independent region proposals. the second is a CNN that extracts a fixed-length feature vector from each region. the third is a set of class-specific linear SVMs.

转载于:https://my.oschina.net/u/3993524/blog/3076141