squirrel last edited by squirrel
I have covered your article about how Bagisto utilized TensorFlow.js for the product search feature.
In the marketplace that we are working on, we will be having pictures similar to the one attached below. Searching this through Bagisto results in keywords like crate wardrobe closet press wooden_spoon. As you see it didn't use keywords like Formica millennium oak which are essential to the image.
My first question:
1- Are these results based on my dataset images or the images dataset that you used in training the feature, or the default dataset from TensorFlow?
2- Could you please guide me on how to train this feature with my customer dataset of images to increase its accuracy?
3- I am thinking of creating a new column in the database to store keywords of the images generated via TensorFlow on upload. This is to help find relevant images on search. Do you think this could be a good idea or there is a better way to achieve acurracy?
Many thanks in advance!
devansh-webkul last edited by
These are default datasets provided by TensorFlow. If you want accuracy then you need to train your model.
For that, I suggest you research machine learning, docs as Bagisto is an opensource ecommerce project.