1
$\begingroup$

How can I build a model which can distinguish between Milk and Phenyl? I want to predict whether a given item is edible to eat or not. If I train a model with thousands of photos of Milk and Phenyl which are labelled, Won't the model get confused between them because they look very very similar? No matter how much I train, I will be getting poor results, won't I? Another such scenario is in the case of salt and sugar.

I want to know if there is a way to tackle this situation. Please let me know if I can achieve better results with any other approach.

Also Please let me know If I'm unclear.

Improve this question
$\endgroup$
2
  • $\begingroup$ It is a good question. Maybe think like that if a human being could simply classify/distinguish them easily or not? For example I could NOT tell the difference between a sugar and salt (if they are almost the grain size, and not branded). I could just tell by tasting them. Then how could you expect a machine learning model to be able to distinguish them? Unless they are in very different bottles (usually) etc. the model might be able to pick it up, otherwise it is impossible, at least to my knowledge. $\endgroup$
    – TwinPenguins
    Oct 15, 2018 at 13:55
  • $\begingroup$ Thank you!! If possible, Please let me know if you found something about this in the near future. $\endgroup$
    – Vinay Varma
    Oct 17, 2018 at 5:53

1 Answer 1

Reset to default
0
$\begingroup$

From the images alone, assuming the links you posted are representative, I doubt a model will ever be able to predict better than random guessing.

There is a chance the model could effectively learn to read the label, but it seems like a long (and somewhat pointless) path to go down, unless you will only be predicting on labelled bottles in the future!

You will need to introduce different information with each sample, perhaps the price, amount of liquid and other such feature. At this point, you will likely want to ditch your CNN for images and instead use a standard classifier, such as Support Vector Machines or a simple feed-forward neural network.

Improve this answer
$\endgroup$
1
  • $\begingroup$ That is helpful, but I'll still keep searching. I'll post as an answer here, if I find something. $\endgroup$
    – Vinay Varma
    Oct 17, 2018 at 5:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.