Training of images in the tensorflow

1

Good evening!

I'm trying to build a neural network in python using tensorflow and its other libraries, my doubt and, I have 4 classes and I need to do a training of them I have many images for each class, I need to use neural network CNN I understand modestly how it works. I need to program mount it in python using tensorflow anyone any examples? Simple in python since I needed to generate the dataset or pb file (protoboof) so I know that it can take the training days to end my biggest difficulty and start python programming from the CNN network.

Does anyone have an example? All that exists on the internet I have seen but almost miguem has tutorial teaching how to generate a dataset file all the tutorials teach with datasets downloaded from websites like ImageNet, Incept, and Mnist. What I need and a simple example of how to generate dataset from my own images

I also read the videos of Iris data (flowers classification) but it also works with information already ready.

Hugs!

    
asked by anonymous 21.03.2018 / 01:43

1 answer

0

You can use this template with tensorflow-hub ( link ). With this template you can perform various manipulations, including including a new dataset. Just run it with the command lines python retrain.py --image_dir "directory path where your folders are with images"

The algorithm will already separate according to the structure of its folders, after training just use an image classified to test new images.

    
23.10.2018 / 16:52