Data-centric AI development proved on PASCAL VOC 2012

In Deep Learning, there are two basic AI solution development approaches - Model- and Data-centric. Model-centric AI development is a conventional way of improving your model. Nowadays, when you start hitting that upper roof of model performance, it might be challenging and expensive to improve results beyond 1-2% on your key metric by tuning the model. However, your success does not rest only on your model's shoulders. There are two crucial components - the algorithm and the data.

So, we decided to check out the Data-centric approach. In short, we tune the data, not the model parameters. Want a peek into the project? Watch the video and see the results by yourself.

Chapters:

0:00 Intro
0:40 TL;DR
1:50 Methodology
4:40 Fixing process
6:37 Neural Network architecture
8:52 Detailed results
11:16 Discussion
12:36 What's next?
13:34 Give us a try!

[hasty]

Let us know what you think!

Leave your thought in the comments or contact us if you want to learn more.