search

jetpacapp / DeepBeliefSDK

The SDK for Jetpac's iOS Deep Belief image recognition framework
Other
2.86k stars 437 forks source link

Build a custom network / Recognize multiple self-learned images #28

Open maxi-mus opened 10 years ago

maxi-mus commented 10 years ago

Is it possible to build a custom network and/or recognize multiple self-learned images simultaneously? If yes, any pointers?

After having studied the available functions in the API it seems to me it's only possible to load a preexisting network. The LearningExample also runs jpcnn_classify_image with a different layerOffset, but doesn't save the prediction values in the newValues array. Why is this done? Do we need to run jpcnn_classify_image and jpcnn_predict for the custom learned object, or is the first part just redundancy?

But my real problem is how I can create a predictor with multiple object, or add objects to the model. From studying the available methods it seems to me this isn't possible, I hope I'm wrong. Any help would be greatly appreciated.

petewarden commented 10 years ago

It's worth posting on the mailing list, I know some folks have been working on changing the predictor to handle multiple classes: https://groups.google.com/forum/?hl=en#!forum/deep-belief-developers

karneaud commented 9 years ago

I'm also interested here. I want to use your framework to detect a custom object i define via image( e.g. label on soda) within another image(e.g. image with said soda)....do you think this is possible?