Hello, I came across your work through the Generative-3D-Part-Assembly project, where your work is used as a baseline. I noticed in the code of Generative-3D-Part-Assembly that the output range of the translation vectors in your pose output, due to the tanh activation function, is limited to -1 to 1. I'm interested in running this code on my dataset, but the span of translation vectors in my dataset is quite large. Could you tell me how you processed your data to accommodate this?
Hello, I came across your work through the Generative-3D-Part-Assembly project, where your work is used as a baseline. I noticed in the code of Generative-3D-Part-Assembly that the output range of the translation vectors in your pose output, due to the tanh activation function, is limited to -1 to 1. I'm interested in running this code on my dataset, but the span of translation vectors in my dataset is quite large. Could you tell me how you processed your data to accommodate this?