简体   繁体   English

如何结合两个快速 R-CNN 模型进行对象检测?

[英]How can I combine two fast R-CNN models for object detection?

I am working on object detection in video, and I want to train two fast R-CNN models on two different types of data (for example RGB and optical flow) then combine these models in one network.我正在研究视频中的对象检测,我想在两种不同类型的数据(例如 RGB 和光流)上训练两个快速 R-CNN 模型,然后将这些模型组合到一个网络中。

I trained the model in one type of data, but I have no idea how can I combine the both type of data.我用一种类型的数据训练了模型,但我不知道如何将这两种类型的数据结合起来。

Any help would be great!任何帮助都会很棒!

What you are looking for are two-stream CNN architectures.您正在寻找的是双流 CNN 架构。 As for example described by Peng and Schmid in Multi-region two-stream R-CNN for action detection .例如Peng 和 Schmid 在 Multi-region two-stream R-CNN for action detection 中描述的 There is a large variaty of implementations but you will need to retrain your two-stream model with optical flow and image data.有多种实现方式,但您需要使用光流和图像数据重新训练双流模型。

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM