search

meguminakao / IGCN

19 stars 3 forks source link

Some inquiries regarding the code #2

Open AlanConnorChen opened 8 months ago

AlanConnorChen commented 8 months ago

Thank you for sharing the open source code! I have a few questions regarding the code:

  1. In the train.py file, "img_label": tf.placeholder(tf.float32, shape=(640, 640, 3)), # target deformation map "shapes": tf.placeholder(tf.float32, shape=(None, 3)), # relative positions Could you clarify how these two variables are stored and represented? For instance, is the target deformation map a mask or a heat map-like image?
  2. In your MICCAI-related paper, you mention using Python version 3.6.8, yet the README indicates Python 3.9 (TMI paper). Additionally, the code in train.py # Specify target organ organ = "liver" appears to be related to the code used for MICCAI. Does this repository currently correspond to IGCN or IGCN+?
meguminakao commented 7 months ago

I missed this question. Sorry for the late answer.

  1. img_label is a two-dimensional mask, a projection image for the initial mesh. This mask can be generated by the simple mesh rendering process.
  2. This open source code includes the IGCN+ framework (TMI version) and a single reconstruction example for the liver shape.
AlanConnorChen commented 7 months ago

No worries about the delay, and thanks for the clarification! The details on img_label and the IGCN+ framework are really helpful.