Fast RCNN trains its own data set

    This blog is mainly about faster-rcnn Train your own dataset （ The data set is VOC Format ）.( This topic focuses on operation ）

    Faster R-CNN Is in fast rcnn Algorithm based on modification . The algorithm consists of two main modules , As shown in the figure below , The first module is a full convolution network for extracting candidate frames (RPN), The second module is based on candidate boxes Fast R-CNN Target detector . The whole detection process is completed through a network .Faster R-CNN Used the so-called attention (attention) Mechanism ,RPN The module tells Faster R-CNN Where to focus .

This document mainly describes how to use faster rcnn Algorithm to realize the training steps of their own data sets , And some solutions to possible problems .

The source code adopts open source code , Links are as follows ：

GitHub - bubbliiiing/faster-rcnn-pytorch: This is a faster-rcnn Of pytorch Implementation of the library , You can use voc Data set format for training . This is a faster-rcnn Of pytorch Implementation of the library , You can use voc Data set format for training .. Contribute to bubbliiiing/faster-rcnn-pytorch development by creating