beijbom

6 ms·

1.
by beijbom 3y ago
(I'm one of the Nyckel co-founders) Chiming in to say that this precisely our observation. The existing ML/DL libraries are not bad as far as those types of things go. In fact, Pytorch is an amazing library IMO. Especially compare
2.
Show HN: Train and deploy an object detector in 10 minutes(youtube.com)
2 points by beijbom 4y ago|1 comments
3.
by beijbom 4y ago
We optimized the training, annotation and deploy infra to minimize the time it takes to bring up a custom object detector. The current version only supports detection of object centers (as opposed to full bounding boxes) and tends to do bes
4.
by beijbom 5y ago
Good question. We have been doing this for almost 2 years now and we still find new players almost every week! It's a bit of a wild west for sure. I can't say what we do different from everyone, but a few things that we focus on:
5.
by beijbom 5y ago
Thanks for your comments fxtentacle! I work at Nyckel. In fact, I'm the "ml guy" at Nyckel. I have a PhD in ML and did some research at Berkeley, but I mostly consider myself a ML engineer. My most recent job was in the self-
6.
by beijbom 5y ago
I'm the ML guy at Nyckel. Great to see all thoughts and comments! Feel free to AMA.
7.
by beijbom 5y ago
Chiming in on the weak labeling question: As of right now, you can use outside libraries like skweak to create weak labels offline and then PUT those using our API ( https://www.nyckel.com/docs#update-annotation ). This would
8.
by beijbom 5y ago
Thanks! I'd love to chat more about your use-case and how we can help you. Drop me a line at oscar at nyckel dot com to set something up.
9.
by beijbom 5y ago
Haha, that's right. Thanks for the shout-out @petters!
10.
by beijbom 5y ago
I'm Oscar, btw. :)
11.
by beijbom 5y ago
Hi Cyril_HN! Thanks for your question. What you are asking for is sometimes called "part of speech" tagging. We currently don't support that but will add it down the road along with more advanced image outputs like detection.