search

yangxuntu / vrd

two models for visual relationship detection
94 stars 28 forks source link

如何从头开始在自己的数据集上训练？ #9

Open CherishineNi opened 5 years ago

CherishineNi commented 5 years ago

  1. 已复现您的模型，现如今想要指定某张新图片(自己的图片)进行测试，请问要如何解决？

  2. 请问如何模仿sg-dataset数据集制作自己的数据集，并完成训练呢？

yangshuo11 commented 5 years ago

same question, if anyone could help with this issue, thanks! @yangxuntu

yangxuntu commented 5 years ago

You need a faster rcnn for detecting the bounding boxes. You need expensive human resources for annotating relationships of objects, then you can build one dataset.

yangshuo11 commented 5 years ago

You need a faster rcnn for detecting the bounding boxes. You need expensive human resources for annotating relationships of objects, then you can build one dataset. thanks for reply!

onexs commented 2 years ago

Is there a file package with a format similar to that in the VRD data set folder for the denoised VG data set