You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi, I'm really interested in COVID-Net CT-2 models, I tried to search through the repo but I couldn't find the model implementation in Python (TensorFlow) so that I can reproduce them with PyTorch. Can you share the original implementation of such models? thank you
The text was updated successfully, but these errors were encountered:
Hi Tran, sorry for the late response. Unfortunately these models were constructed using a machine-driven design strategy in which the original model is modified automatically. The downside of this approach is that it produces a TensorFlow meta graph file for the modified model, but not the source code for constructing the modified model.
To convert the model to PyTorch, your best bet would probably be to view the graph in TensorBoard - this allows you to see the layers, connections, input/output shapes, channel counts, etc. There is also a somewhat high-level diagram of the architecture in the CT-2 paper which can help with the general structure, and this diagram also includes the channel counts for the various layers.
Hi, I'm really interested in COVID-Net CT-2 models, I tried to search through the repo but I couldn't find the model implementation in Python (TensorFlow) so that I can reproduce them with PyTorch. Can you share the original implementation of such models? thank you
The text was updated successfully, but these errors were encountered: