Message Boards Message Boards

0
|
2209 Views
|
4 Replies
|
3 Total Likes
View groups...
Share
Share this post:

Standard configuration for ImageIdentify[ ]?

Posted 3 years ago

As per https://reference.wolfram.com/language/ref/NetModel.html uses a lot of different configurations of neural networks and underlying training datasets.

However, I wonder which is the Neural Net and the Training Data Set being used for ImageIdentify[ ] without any parameters? My best guess is, that it is ImageNet based, but I would be grateful for more precise information. Thanks in advance, Francis

POSTED BY: Francis Hunger
4 Replies
Posted 3 years ago

Thanks Rohit.

POSTED BY: Francis Hunger
Posted 3 years ago

Hi Francis,

The Wolfram NN repository page for ImageIdentify says

over 3 million training images and over 4,000 classes of objects (not publicly available).

The network architecture is available on the NN repository page, you can compare it to VGG19 to see what is different. You could try running the ImageIdentify NN against the ImageNet images and compare performance with VGG19.

You could also try contacting Wolfram Support and asking for more information on the training set.

POSTED BY: Rohit Namjoshi
Posted 3 years ago

Thanks Rohit, that explains a lot already. Is it right to assume, that Imageidentify was trained on ImageNet (+potentially other ressources)? There seems to be no statement about it. When using Imageidentify for scientific research it would make sense to know these details in order to see how it differentiates e.g. from VGG19 on ImageNet.

POSTED BY: Francis Hunger
Posted 3 years ago

From the documentation for ImageIdentify under Properties and Relations

The neural net used by ImageIdentify can be accessed using NetModel:

net = NetModel["Wolfram ImageIdentify Net V1"]

Details here. The training data is not public.

POSTED BY: Rohit Namjoshi
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract