Hi guys i really need your help. I recently made an app that would be able to classify the type of soil one has using a photo taken by the user. I used the personal image clasifier on the MIT and used photos of different soil and calssified them with their soil types and trained the model to identify the soil type using a percentage level..but now my accuracy of the classification is not enough and i want to improve it and i don't know how to
Thank you
Using a perceived percentage level for classifying soil type even with a human estimating percentages of content is not very accurate. You might want to be satisfied with the present capabilities of your soil classifier.
This discussion might help to allow you to sieve the soil and extract particle size.
Also consider more examples in your 'training' set of images.
Hi Steve thank you
I have considered trying to put more data in the training set but it’s just not enough data because I take the photos from the internet (and categorise the photos with each soil type) and I have tried contacting agricultural companies for more accurate soil data etc but they are just not responding.
The discussion you recommended I skimmed over it and feel it’s a bit too much for my knowledge as I’m an only just a starter
Regards