Yes, my feeling is that we're in an "in between" situation: - the promise of AWS et al. is that deep learning is available to every developer ; this is true on

Yes, my feeling is that we're in an "in between" situation: - the promise of AWS et al. is that deep learning is available to every developer ; this is true on a purely technical basis, but without appropriate knowledge and experience, constituting a curated database is really hard. - CV expertise is a little bit overrated nowadays, as I believe key success factors for a CV project is not the technical expertise of the algorithm wizard, but rather the rigorousness of the project management (data science is 20%, the rest is 80%). I may be wrong, but both approaches lack the same thing: a way to curate the image dataset — and get those 80% of the work done without hassle. Once these 80% of data management is done, the main difference between the "CV amateur" and "CV expert" is the ability to push the performance of the algorithm. But maybe we just lack a "Postgres for images" that would make these 80% easier, after all :)