network design

[2a3b4c]

It's an excellent work. I'd like to ask if the network design of your work is similar to that of EasyControl?

[innerway-xq]

Hi, thanks for your interest in our work!
Although both approaches involve conditional generation, they tackle different tasks. EasyControl focuses on overall geometric control (e.g., depth, canny, subject mask), while InstanceAssemble targets layout-guided image generation, emphasizing localized layout control — generating images that follow the spatial and semantic information of the given layout.
Because the tasks are different, the network designs also differ. EasyControl handles multiple conditions through attention mask constraints, whereas InstanceAssemble introduces the Assemble-Attention mechanism to model interactions among instances. Please refer to our paper for more details.