Need suggestions, directions : attempting to create a simple model to recognize backyard trees

I am a relative beginner, trying to get familiar with HTM (by reviewing various jupytr notebooks shared on the forum - using a docker image).

Looking forward to suggestions, directions, or leads.

In order to make the learning process more grounded, I am doing following:

  1. Collecting images of trees in my backyard.
  2. Creating a layered image (using paint) to mark stem, leaf, flower etc.
  3. Using chatgpt to trouble shoot when stuck.

What I anticipate, there are about 15 trees (such as: fig, guava, moringa etc), over the summer they will be having a growth spurt. In my limited understanding a model when given an input of image should “predict the tree” easily.

Similarly, when using a video as input - should be able to label / create a bounding box?!

With encoding spatial information of individual tree, i.e. where they are relatively located on a satellite image of backyard - should make it interesting?

Thus looking forward to thoughts, on what all should one consider. For example:

  1. Creating a data set?
  2. Structuring the encoder, pooler pipeline?
  3. How to make it functional / utilize it?

Thank You.

Example of Image and annotating stem layer.

1 Like