0

Is there has any method to train Tensorflow AI/ML that I focus on detecting background of image more than common objects?

I'm newbie to ML field, but was assigned to do job that make an application which can detecting on showroom image/places and detecting on the floor, wall then find out what is the material/ceramic/marble/etc. product they are.

Example: This is showroom picture,

enter image description here

the wall and the floor of showroom are using this product material

enter image description here

  • Is it possible to do something like I described?
  • How to start with?
  • If I don't want to install Tensorflow into my computer, is there a service that can make a model to use in the device? (my goal need to use the model in Android device)
  • What method/type of ML should I approach 'Classification' or 'Object Detection' or other else?
Natta Wang
  • 119
  • 3
  • Hi and welcome to this community! Please, ask just **one question per post**, otherwise, your post could be closed as too broad. – nbro Nov 19 '19 at 15:23
  • @nbro sorry, but what will you think if I just flood the board with many of questions or just gather them all into one post. So I can ask and clear my minds from self-learn questions as a newbie like me? Hope you understand the blinds people who come into new field of knowledge and try self-learn first but there's ton of technical terms that can't be cleared understand from read them once. – Natta Wang Nov 20 '19 at 09:32
  • You should ask one question per post, if you are either a newbie or not. Furthermore, these questions is off-topic here. – nbro Nov 20 '19 at 12:32

1 Answers1

4

Trying to address all the questions asked in the end in the same order

  • Most definitely possible.

  • I would say its best you approach this with segmentation to start with.

  • Just use a free GPU runtime notebook service such as Google Colab or Kaggle Kernels. But you would not directly be able to integrate with the device, you'd have to keep moving input and output from your drive (on Colab). There might be a better service for the needs described, but this is the best I know on this.

  • Your background can be segmented and the segment can work on transforms such as maybe convolutions or affine transforms to be able to get the relevant information regarding the background.

Hope this was helpful!

ashenoy
  • 1,409
  • 4
  • 18