I have an image classification problem but how do I start?

Can you tell me how I can start building an image classification model? I do not know how to start.

Thank you.

Hi Tosca,

You can start with having a look at the image classification cheat sheets at Knowledge Center: “Cheat sheets to solve AI problems on the Peltarion Platform

Also, the best way to learn is to try out! You can try this car damage assessment image classification tutorial: “How to classify car damages with deep learning using Peltarion

If you don’t want to read, you can also watch: “3. Image classification - Car damage tutorial | Peltarion x Adalo - YouTube

When you are done, check out the other tutorials for image classification: “Deep learning tutorials | Knowledge center | Peltarion

Happy modelling!
Begüm

1 Like

Hi!

Adding to Begüms nice answer, that constructing the dataset can sometimes feel trickier than it actually is:

Your dataset should consist out of a balanced amount of images (if you have two classes then it’s 50/50, and so on).

If you have a small dataset you need to consider standardizing how the object is represented in the image (angle, close-up, season etc.), if you have a larger dataset (1000+ per class or so) this is of less importance as you will have better representation of this in the dataset.
But this is very much a case-by-case thing, so start with what you have, train a model and see how it performs in the Evaluation view (use the confusion matrix, focus on the worst parts).

The dataset format requirements you can find here.

But to make it simple: Organize your images in 1 folder per class, create 1 zip file with all the folders in (they should all be at the the root of the zip-file, no sub-folders or similar), then just upload the zip-file and go for it!

Good luck!

1 Like