search

1348722633 / TensorRT_FasterRcnn_Windows

2 stars 1 forks source link

if i use my own datasets, how to set the "pixelMean[3]" in FasterRcnn::prepareImg function? #1

Open brucelee78 opened 3 years ago

brucelee78 commented 3 years ago

hello , thank you open your code~ i have some questions:

  1. if i use my own datasets, how to set the ‘pixelMean[3]’ Array in ‘ FasterRcnn::prepareImg’ function?
  2. which faster rcnn code trained the model， can compatible this tensort code，Do you have a recommendation？

thank you！

1348722633 commented 3 years ago

I think using your own data set without changing this average should not have a big impact. You can try to calculate the RGB channel average of your own data set, but you also need to do corresponding data preprocessing during training.  Regarding the training code of Faster RCNN, you can go to github to find the official caffe code to try. It is possible that a custom operator is needed when RPN outputs the proposal.

------------------ 原始邮件 ------------------ 发件人: "1348722633/TensorRT_FasterRcnn_Windows" <notifications@github.com>; 发送时间: 2020年9月27日(星期天) 晚上6:00 收件人: "1348722633/TensorRT_FasterRcnn_Windows"<TensorRT_FasterRcnn_Windows@noreply.github.com>; 抄送: "Subscribed"<subscribed@noreply.github.com>; 主题: [1348722633/TensorRT_FasterRcnn_Windows] if i use my own datasets, how to set the "pixelMean[3]" in FasterRcnn::prepareImg function? (#1)

hello , thank you open your code~ i have some questions:

if i use my own datasets, how to set the ‘pixelMean[3]’ Array in ‘ FasterRcnn::prepareImg’ function?

which faster rcnn code trained the model， can compatible this tensort code，Do you have a recommendation？

thank you！

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.