Two question about evaluating own image!

[jxylon]

Now I extract bottom-up-attention features on my own image without image scene graph.

1. If I want to use image scene graph, What should I do?
2. Is your modelid660066 trained use rela or not?

Thanks!