188 karma · joined December 31, 2014
That being said, I would like to see a platform for discussion from BOTH parties instead of blogposts with no real communication.
Since COVID people are using it less, but consumption is not all that low surprisingly. And yes it's not 100% compared to a chemical test I know. Same tech but better use than facial recognition.
I stopped using gpu's, "Vectorized inference isn’t bad at all!". This soo much, I was blinded with gpu speed, using tensorflow builds with avx optimization is actually pretty fast.
My discovery:
+ Stop expensive GPU's for inference and switch to avx optimized tensorflow builds.
+ Cleaned up the inference pipeline and reduced complexity.
+ Buying compute instance for a year or more provides a discount.
- I never got pruning to work without a significant loss increase.
- Tried spot instances with gpu's that are cheaper. Random kills and spinning up new instances took too long loading my code. The discount is a lot, but I couldn't reliable get it up. Users where getting more timeouts. I bailed and just used cpu inference. The gpu was being underutilized, using cpu only increased the inference to around 2-3 seconds. With the price trade off it was a more simpel,cheaper and easier solution.
And I created a reverse image search to look up pill data: https://play.google.com/store/apps/details?id=be.harmreducti...
It was my first project getting into machine learning.
It's made with 2 models, a Retinanet model for localization and a triplet loss model for generating image embeddings.
Those are then compared in an Annoy database and voila.
Feedback is appreciated. Thanks
Custom Android tablet with speech to text: getting my deaf grandmother out of social isolation.
Converting an old school moped to an electric high performance version: On hold due to COVID-19 ...
Also that is very depending of your local laws. I know for a fact that this feature is currently being developed by a harm reduction group of the EU.
Surprisingly I would much rather have it owned by the EU.
Its still very alpha, but could make it easier for people that don't have access to kits.
It's not chemical validation tho.
There are compounds like NSI-189 that are going towards this path, but they come up short in trails.
If you click it, it will probably point to the play store to install it like any other ad.
Change the copy to "install more of my apps" and they will allow it.
Yes since they employed lots of manual labor, many false positive have hit me as well. And it's very frustrating.
But this is just another empty "Lets hate on Google". Following the guidelines this IS a deceptive ad.
Simple reply them in the policy form and they'll get back to you why within 2 days.
But as an indie dev that has 50% of all apps currently suspended for bogus reason, I'm frustrated, angry and looking for a more stable platform.
Article: https://www.hln.be/regio/antwerpen/rechter-straft-it-special...
Check out the image in the article. They attached keyloggers and sent the strokes to the box. Saving them and once in a week dump them over to a car in the parking lot.
The original article is great, but the guy was really not putting any effort into it.
Going to check out how to process my model from keras to use with this tool. Doesn't like that complicated.
Gonna give this a shot :)
Great work !