Image Processing - Keras

Dear all,

i am having a little bit of experience in KNIME, but not with Image-Processing…

What am i doing?
I am reading a CSV file and a png Image. The Image is the air route from a drone. The CSV file is full of alle information - regarding altitude hold, GPS coordinates and so on… with these information i am trying to develop a Deep Learning network with Keras. So that in the end - Keras can decide, how good the pilot has flown the maneuver.

My questions are:

  1. Is that even possible in KNIME?
  2. Can i get the importiert image coloured?

thanks in advance

1 Like

Hi @buechertasche123,

I’m not entirely clear on your use case, but I’ll try to speak to the topic a little.

Combining CSV data and Image data to be used as inputs for a Keras network is definitely possible in KNIME. Here’s an example that uses images as inputs to a deep learning model.

As to your question about reading in colored images, yes that is possible, the image reader will keep all of the color channels by default.

What extra information do you hope to extract from the image that wouldn’t be included in the gps, altitude, etc data in the csv?

Hi Corey,

super thanks for your reply!

How can make it more specific? Or could you please tell me what is unclear?

As i thought! Problem is - the image is black/white and not coloured.

The 2D Speed and the vertical speed. This is how it looks like:

regards

1 Like

Hi,

i solved it. Problem was that Apple has its own interpretation of png files… The Screenshot was saved as png (apple style) and the KNIME image reader wasn’t able to read it, except the x and y axis. The coloured part z axis was missing.

How to solve?
convert the png mac Screenshot into a JPEG.

It would be great if KNIME could prevent a solution for that :slight_smile:

greetings

1 Like

Thanks for sharing that solution!
I work on a windows laptop and wouldn’t have thought to try that.

I’m curious if the png screenshot was just oriented differently, in which case the Dimension Swapper might be useful as a way to fix it without converting file types.

Hi Corey,

  • when i am using “image properties”, i am always getting the same amount of dimensions, doesn’t matter if mac png or windows png.
  • As far as i know the z-axis is the “coloured part of the image”.
  • When i am using “Auto Crop” the image gets coloured - by mac png, but unfortunately it shows the image in the wrong size.

I have tried the “Dimension Swapper” as well, but it’s not working.

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.