Thanks for your great work!

However, I have some problem about sketch rendering. I'm wondering how the line is perturbed with the endpoints alone are applied with noise.

Is there any code I can refer to?

Hello,

I've successfully obtained the configuration details for several models, but I'm unclear on how to utilize them for converting 2D sketches into 3D CAD models. Specifically, I'm puzzled about where to input my 2D sketch and how the regNet_300k model can generate a 3D CAD model from a sketch, especially when dealing with unusual shapes.

Here are the input and output specifications for each model:

embedSNet Input and Output Names and Shapes:

• Signature: serving_default
  • Input Name: input_tensor, Shape: (None, 256, 256, 1), Dtype: <dtype: 'float32'>
  • Output Name: output_recons, Shape: (None, 256, 256, 1), Dtype: <dtype: 'float32'>
  • Output Name: output_code, Shape: (None, 256), Dtype: <dtype: 'float32'>

gpTFNet_250k Input and Output Names and Shapes:

• Signature: serving_default
  • Input Name: dec_padding_mask, Shape: (1, 1, 1, None), Dtype: <dtype: 'float32'>
  • Input Name: look_ahead_mask, Shape: (1, 1, None, None), Dtype: <dtype: 'float32'>
  • Input Name: inp, Shape: (1, None, 256), Dtype: <dtype: 'float32'>
  • Input Name: tar, Shape: (1, None, 256), Dtype: <dtype: 'float32'>
  • Input Name: enc_padding_mask, Shape: (1, 1, 1, None), Dtype: <dtype: 'float32'>
  • Output Name: res_group_label, Shape: (None, None, None), Dtype: <dtype: 'float32'>

regNet_300k Input and Output Names and Shapes:

• Signature: serving_default
  • Input Name: gp_S_map, Shape: (1, None, 256, 256, 1), Dtype: <dtype: 'float32'>
  • Input Name: gp_ND_maps, Shape: (1, None, 256, 256, 4), Dtype: <dtype: 'float32'>
  • Output Name: base_curve, Shape: (None, None, 256, 256, 1), Dtype: <dtype: 'float32'>
  • Output Name: face_map, Shape: (None, None, 256, 256, 1), Dtype: <dtype: 'float32'>

Given these details, could you guide me on the correct process for feeding my 2D sketch into the system and using regNet_300k to create a 3D CAD model? Any insights or examples would be greatly appreciated.

Thank you!

Hello,
Firstly, I appreciate your commendable work on this project.

I've successfully set up the environment and verified it by running python train_AES_SG.py without any issues. Upon reviewing the supplementary material provided at this link, I noticed the use of Fusion360 for testing. However, I'm uncertain about how to leverage your pretrained network within Fusion360.

Could you kindly provide detailed instructions on this? It would be of immense help.

Thank you!

2023-09-01 08:06:15.849558: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer.so.6'; dlerror: libnvinfer.so.6: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /usr/local/nvidia/lib:/usr/local/nvidia/lib64
2023-09-01 08:06:15.849640: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer_plugin.so.6'; dlerror: libnvinfer_plugin.so.6: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /usr/local/nvidia/lib:/usr/local/nvidia/lib64
2023-09-01 08:06:15.849653: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:30] Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.
2023-09-01 08:06:18.093492: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer.so.6'; dlerror: libnvinfer.so.6: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /usr/local/nvidia/lib:/usr/local/nvidia/lib64
2023-09-01 08:06:18.093573: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer_plugin.so.6'; dlerror: libnvinfer_plugin.so.6: cannot open shared object file: No such file or directory; LD_LIBRARY_PATH: /usr/local/nvidia/lib:/usr/local/nvidia/lib64
2023-09-01 08:06:18.093588: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:30] Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.

Hello,

I've successfully set up and executed the Fusion360 Python plugin available at AutodeskAILab/Fusion360GalleryDataset. However, I am encountering two challenges:

Dataset Compatibility: I'm unsure how to modify the Fusion360GalleryDataset so that it becomes compatible with the format you require for your plugin or tool. Could you provide a step-by-step guide or outline the necessary transformations?

Command Sequences to CAD: I need detailed instructions on how to take the command sequences from the dataset provided and use them to generate the corresponding CAD models in Fusion360. A walkthrough on this process would be greatly appreciated.

Hi，I'm currently working with your source code and I came across the trained_model.h file. In that file, I noticed a reference to "util_functions.h." I'm wondering if this header file is something you've written yourself or if it belongs to a specific library.

Could you provide more information about the origin and content of "util_functions.h"? It would be helpful for me to understand its purpose and whether it's part of your codebase or an external dependency.

Thank you for your time and assistance!

Thank you for sharing the code for the wonderful paper! Could you please tell me the following? I cannot download the ‘Training data’ and ‘checkpoint’ mentioned in the README of the dataAndModel directory because the links are not working. Could you please tell me how to download them?

Hello,

I hope this message finds you well. First and foremost, I would like to extend my gratitude for making these training scripts (train_AES_SG.py, train_grouper.py, train_Regressor.py) available. I have been diligently working to train them on my end.

Despite achieving a relatively low loss, I have observed that the prediction results still deviate significantly from the Ground Truth. I understand that you have kindly provided a checkpoint file, available at this link. Upon inspection, I suspect that the content might be the test results instead of the network weights, though I stand to be corrected.

I am reaching out to kindly request if it would be possible for you to share the trained network weights resulting from all three scripts mentioned earlier. Having access to these would be immensely helpful and would aid me significantly in understanding the expected outcomes of the training process.

Thank you very much for your time and assistance in advance. I am looking forward to your positive response and am more than willing to provide any additional information that might be required.