Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling

meetps / tf-3dgan

Tensorflow implementation of 3D Generative Adversarial Network.

https://meetshah.dev/gan/deep-learning/tensorflow/visdom/2017/04/01/3d-generative-adverserial-networks-for-volume-classification-and-generation.html

MIT License

284 stars 79 forks source link

Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling #15

Open lkffight opened 7 years ago

lkffight commented 7 years ago

@meetshah1995 Are there any links between the input 2D images? 2D image to 3D image contact is what? Thank you

lkffight commented 7 years ago

@meetshah1995 Did your results reproduce now? I use your code to get the result i do not know what? What is the specific format of your code input and output? Thank you

DJxuelei commented 4 years ago

Hi ,I think the input may be A random normal distribution as you can see in the paper. it can be get by code: z_sample = np.random.normal(0, 0.33, size=[batch_size, z_size]).astype(np.float32).