Data is vital for realizing machine leaning algorithms into industrial products — as sensing the strong feedbacks from the attendees from my 50 min talk in NVIDIA GTC 2018. Here, we summarized how and where to get the quality training data for advancing deep learning, motivated by our own research and those collaborations from the industrial partners.

Feel happy to have successfully that in the topic of “Investigating Data Augmentation Strategies for Advancing Deep Learning Training.” We show 4 major strategies to collect quality data in an efficient and effective way for advancing deep learning training. Some industrial applications and advanced researches are illustrated in the presentation as well. It’s more exciting that we have a fully packed attendees (overflowing on the floor) in a large lecture room (>150?).

Apparently the deep learning practitioners care a lot about DATA, which is vital in the real applications. The sides can be downloaded here.

The sides can be downloaded here.