I'm interested to understand the biggest pain points within your machine learning and computer vision workflows.

For me personally it's the creation and labelling of datasets.