At Flytta, We tend to bring information needed for a person to his fingertips, such that he can find what he wants in simple clicks.
One of such a solution is this, where we extracted the items from images that were uploaded from homes which were vacant.
These images were passed through a DL pipeline where we capture the items in it and use it to query the customer's input.
Example of this image, if a person is looking for a house with chairs and dining table, this house information is shown.
We did a transfer leaning from Imagenet using our custom dataset to train the model and we also repetitively do manual correction to tune it.