Spaces:

torfasonc

/

Accord_or_Civic

like 1

Sleeping

Can you please explain to me from where did you get the data from ?

Hey Ravi! I grabbed all the images through DuckDuckGo (which did include a fair bit of junk images). I then cleaned up the dataset after an initial training.

TYSM!

while doing data cleaning, Jeremy metioned that we can build a model on the raw data and use some some function to extract top 10 losess and clean them manually.

my question :-
Is there any way to automate this, as we are just doing a manual trial and error on some top 10 images when there thousands of images in the dataset ? or is it so specific to the data at hand that this is the only way to go about it ?