So dataenginners are as important as data scientists
From the replies: if you're going to start a ML services startup, make it about data collection and labelling. This is main pain point, and where the most value can be unlocked.https://twitter.com/fchollet/status/1074365753053200384 …
-
-
This Tweet is unavailable.
-
-
-
-
This Tweet is unavailable.
-
Sounds like something
@sequencework can do
End of conversation
-
-
-
This Tweet is unavailable.
-
For generic images, large data sets already exist (imgnet, coco) & used w/ transfer learning to the problem — the difficulty is when you have considerably different domain, eg medical images specific to some pathology, then the labeling is domain specific (eg not bounding boxes)
End of conversation
-
-
-
If your looking for labeling help especially for NLP, check out prodigy from the folks
@spacy_io. Works with image too!!! -
This Tweet is unavailable.
- Show replies
New conversation -
-
-
I am increasingly beginning to believe this. This is an are where the major cloud vendors (with all their computer) do not have as much of an advantage (at least not yet). Data collection / labeling is hugely painful and not yet 100% figured out for generalized use cases
-
This Tweet is unavailable.
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.