RE: A Beginning Journey in TensorFlow #6: Image Augmentation and Dropout

Sean_Miller — Sat, 23 Nov 2019 15:48:45 GMT

Great blog series! Hope to see more.

My big question is, after I run code to train a model, how to I use it with my BBAI classification.cpp code?

The classification example has close to a 1000 objects it runs through it appears. The default code takes it down to just 10 its interested in for the demo - although I think it is still processing against all 1000. I don't know where the model data actually sits, so I'll try to figure this out today.

The magic of all this is to train some things like "door open, door closed" and then be able to get models on an embedded device that can help you out in life. I'll post back here if I can figure it out.

-Sean

RE: A Beginning Journey in TensorFlow #6: Image Augmentation and Dropout

genebren — Sun, 20 Oct 2019 16:30:34 GMT

Frank,

The augmentation concepts seem very powerful. Covering a range of angles and sizes seems like a great way to stretch your images to improve your model. Images that are going to be recognized will not always be oriented straight up and down or at a given size, so I can see how this might improve accuracy.

Well done!

Gene