Thanks for providing such great job, and I still have two question:
1.In the Data Construction section of the paper, I saw that generating 2D training data requires 3D models from the Mixamo material library, which is for obtaining more accurate line segment endpoints. If my training data only has 2D images, as long as I can find the accurate line segment endpoints in the images and process them into the same training data format, can I also use this method to train the model?
2.Will you open-source the code for converting 3D models into the provided training data style in the future? Or could you explain in more detail how to convert 3D models into the training data style?
Thanks for providing such great job, and I still have two question: 1.In the Data Construction section of the paper, I saw that generating 2D training data requires 3D models from the Mixamo material library, which is for obtaining more accurate line segment endpoints. If my training data only has 2D images, as long as I can find the accurate line segment endpoints in the images and process them into the same training data format, can I also use this method to train the model? 2.Will you open-source the code for converting 3D models into the provided training data style in the future? Or could you explain in more detail how to convert 3D models into the training data style?