Example code to measure the performance of tf.data API

The example code predicts the class label of input image by using VGG19 network

For experiments

You should download pre-trained VGG19 weight vgg19.npy

1. If you want to use tf.data API, pipe_lining option to True
2. If you want to read data with main memory, in_memory option to True
3. If you want to read data without main memory (on disk), in_memory option to False

Experimental Results

I measured the average time for predicting class label.

GPU : Nvidia Tesla K80

CPU : Intel(R) Xeon(R) CPU E5-2686 v4 @ 2.30GHz

Batch size : 32

	With tf.data API	w.o tf.data API(using feed_dict)
in_memory	0.3459s/batch	0.3410s/batch
disk based	1.4656s/batch	4.0910s/batch

Comments

If you have any questions or comments on my codes, please email to me. [email protected]