Training Chip and Target Image format in TensorFlow

Question

I am trying to build a Land Cover Classification model for Sentinel Image. The Image Channels(bands) I am using are 32-bit float.

I need to understand how to best format the Image data, both the chips/patches for training and the Target Image for Classification. I have few questions?

Do I need to convert my Original Image and Training Chips from 32bit to other depth?
Do I need to ensure that both the training chips/patches and target have same depth (either 32bit, 16bit or other)?
Do I need to resale my data? I saw some papers where data was rescaled between 0-1 or 0-255?
Does data depth effect the performance of learning and predicting?

Many thanks.

Maz

score 0 · Accepted Answer · answered Oct 09 '19 at 14:50

0

The best precision to use on a PC is float32 for many reasons like, more precision makes calculation more accurate which is better, but somehow float16 is slower than float32 on PC(I don't remember why) and float64 is unusable slow on regular machines.

So

You usually need to use float32 as input anyway. So if it's float32 in the first place then just use it like that.
You do, but I think they all will got converted to ther desired precision during fit or predict for keras. It's in $HOME/.keras/keras.json.
I don't think it's a need but std centered rescale helps convergence, though, google always simply rescale to -1 to 1.
It does, but as I said, more precision gives better accuracy but it slower.

answered Oct 09 '19 at 14:50

Natthaphon Hongcharoen

2,244
1
9
23

1

Found another helpful discussion: https://stackoverflow.com/questions/46613748/float16-vs-float32-for-convolutional-neural-networks – Hussain Oct 15 '19 at 17:52
Yeah, but as 16bits runs slower than 32bits in some machines so you need to select what fit your machine the best. – Natthaphon Hongcharoen Oct 16 '19 at 08:29
There's also another data type call `quantized float` and `quantized int`. – Natthaphon Hongcharoen Oct 16 '19 at 08:32
Usually we use both `float16`, `qfloat` and `qint` in `arm` gpu like phones or IoTs, for these machine they are way faster than `float32` – Natthaphon Hongcharoen Oct 16 '19 at 08:33
For `float64`.... well if you want to train it using gpu I think increase model size is a more efficient in both accuracy and time – Natthaphon Hongcharoen Oct 16 '19 at 08:35

Training Chip and Target Image format in TensorFlow

1 Answers1