This is the Remyx!
Hi all,
At Remyx.ai, we've been developing the easiest way to build custom ML models.
See our early prototype below:
We've written a post describing the Remyx Model Engine, check it out: https://smellslike.ml/posts/the-remyx/
Trending on Indie Hackers
no idea. but the gif is powerful
if you ask me in a business context, then I could answer
We hope to start by enabling more computer vision applications.
Consider developers using this camera: https://www.luxonis.com/applications
or building AR/VR applications like these: https://ar.snap.com/lens-studio
Often access to data and expertise is a limitation and we hope to simplify this with a trove of billions of images and autoML.
you say: But the context/environment will vary and we want to help users adapt quickly.
That is the no 1 lever for traction/growth. Make it "plug and play-able" as the space matures. Context, Environment are part of the lego blocks.
If I had twigged 5 real life applications, then in my next business interactions, I would be showing to "create possibility" in the mainstream audience. When the audience are in the DKDK (dont know that dont know), first step is creating the realm of possibility. Make that easy and you have some traction.
Totally agree, we hope to showcase more examples on YouTube, Hackster.io, DevPost, and putting together some good docs. It could make sense to compile a short guide at some point.
AR.snap looks incredible. Now i have to work out how to leverage off it in real-world without marketing noise.
Would it solve / or assist and mobile data capture business problems (e.g. smarter than citizen scientist or snapsendsolve). I know this is a pain point as data is crowd sourced and data build is public. Not seeing anything good in this space yet
We would have liked to have a V0 out in time for this contest: https://snaplensathon.devpost.com/
But they'll have another and we think many ppl will want to build vision pipelines with a custom model.
excellent idea. Oh its been and gone. Thanks for info. Will look into.
What if the audience doesnt use snapchat. LIke real businesses with little time for such play.
From the standpoint of our engine, its all the same but some developers will target mobile devices to sell apps in the store, others will want to target microcontrollers for IoT applications like in this arduino post:
https://blog.arduino.cc/2021/07/13/droop-there-it-is-is-a-smart-irrigation-system-that-uses-ml-to-visually-diagnose-drought-stress/
I worked with the OAK-D camera to build perception pipelines for autonomous mobile robots to be deployed in different customer warehouses.
Then we expect others will have use for a hosted endpoint they can hit infrequently for these tasks.
incredible. Is there a "tour" or 'simulation' of the work done 'OAK-D camera to build perception pipelines for autonomous mobile robots to be deployed in different customer warehouses.'
Like as mundane but helpful as A or as critical and life saving as B
A- many countries, have doggy parks. But the dog poop bags are always missing. when bag pouch has a refill/reorder point, and its sent back to council
B- live in an area where timber slash is a real problem. Clogs up the river and was a major cause of the recent Esk Valley flooding in Hawkes bay, NZ. Could citizens plant their own watchers, so they dont have to walk up every stream branch regularly to keep an eye out.
We have been adding demos of that flavor in our twitter for content marketing.
https://twitter.com/remyxai
Thinking a "model-of-the-day" campaign may be interesting, what do you think?
you say: The MJ discord idea is interesting, I've thought a lot about how they are generating virality in the shared channels and helping others learn.
I even subscribed just so i could participate in a feedback survey and would happily pay it forward my fast CPU minutes for a month, if there was a way. Building together in public is big hook. aka chatgpt plus launch approach. (I am also a chatgpt plus subscriber and i hate subscription models for micro). But it is revolutionary, so I did after 2 days of use.
Twitter is not the platform for driving community engagement at the level of contribution, build. Citizens need context. side by side comparison - showcase.. Screen realestate on device is a real constraint. Twitter is dialogue on a point, or look at me. It does those interactions very well. Content marketing typically has low return as fails as it is not clear 'what intereaction' are you wanting to generate, encourage, nuture.
Well put, we've often found that twitter follows come after endorsement by a figure within the community. It can be tricky to straddle the line in communicating a technical product to a broader audience you expect would want to use it.
Yep! I would do it on web platform too, like MJ has. Showcase on discourse as many are twitter withdrawers
or solve this problem with a Model. Circular Economy ran those to solve the material challenge to switch to a waste free, circular economy
another example, this group open-sourced a detector for parking cones:
https://github.com/mkrupczak3/Coneslayer
You can stuff a dozen models onto these devices:
But the context/environment will vary and we want to help users adapt quickly.
The MJ discord idea is interesting, I've thought a lot about how they are generating virality in the shared channels and helping others learn.
Maybe we can do something similar with the prompts, there is still much room for companies to remain differentiated using this engine because we plan to ingest data.