Discussion

Ask HN: Dataset Cleaning Tips

Hello! I am building a custom dataset in the style of CIFAR-10, so I have about 60k images. As of now I'll probably be cleaning the dataset manually be deleting non-relevant images, cropping, etc, in addition to automatic preprocessing like resizing images and so on.

Do you have any tip or best practice for the manual cleaning up? Is using something like Amazon Mechanical Turc doable/viable?

Thanks!

3 points ratsimihah 7 years ago 1 comments

Discussion

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton