Open teolemon opened 6 years ago
@teolemon I would like to take up this issue. I have already started working on it . To start with i'm planning to take around 800-900 images and classify them in two broad categories i.e bottles/cans and others
I will start will manual data annotation and then use the classifier to further annotate the data. I'm already half way through it
Here is the github repo of the classifier i was taking about : https://github.com/vasu-dev/packing-classifier
the results are not very good as i just used 881 samples due to memory restrictions. I have trained the model with data argumentation
A better approch to this problem would be first Recognize packaging material (https://github.com/openfoodfacts/openfoodfacts-ai/issues/23) and then used it to find the packing type . I would love to combine them and work on it as my GSoC project.
I'm working on it!
Right now I'm looking to the datas and figuring out how many labels we are going to need to be pertinent.
I'll look to your work @vasu-dev. Did you find why it doesn't work?
I'd love to work on this issue, please assign it to me if nobody is working on it. @teolemon, Should I start working on it, or I have to wait until it gets assigned to me?
@ademenet Are you still working on it?
I started looking at this - doe something have curves and straight lines: cylinder; corners and straight lines? Box
But after a bit of thought, #55 seemed an easier problem to solve - perhaps could be used as a validation for training.
What
Part of
97