0
$\begingroup$

I am new to deep learning, I hope you will lead me because I have been stuck for a week. I am trying to build a model for identifying a single object in the image. So, I made my custom dataset, which looks like this:

  1. I have 3 objects e.g dog, cat, cow. (So only 3 pictures which are being pasted)
  2. Object is randomly placed around some (on top of it) background picture.
  3. My dataset has three folders -> Train (has images (jpgs) and labels (txt)). Test with similar and Val with similar.
  4. In my labels, I save the class (0,1,2) and where the object is pasted, so it's coordinates.

Now, I want to do my CNN architecture.

I want the model to have 2 outputs:

  1. one for class probabilities (same size as the number of objects chosen, i.e. 3)
  2. another for bounding box regression (of size 4, due to the box coordinates x,y,w,h)

How do I go on in creating the model? What if I want regularization. Is there an online repo about this? or an article. I can't find something which does training for one object.

Improve this question
$\endgroup$

0

Reset to default

You must log in to answer this question.