alexcnwy

18 ms·

151.
by alexcnwy 7y ago
Super interesting, thanks for the reply! :)
152.
by alexcnwy 7y ago
Yes, of course! Essentially centroid tracking works like this: 1. Detect objects (e.g. using TensorFlow) in each video frame 2. Initialize an incrementing ID for each object at the first frame 3. Compute the centroid of each
153.
by alexcnwy 7y ago
This is an interesting and well-written post but in my experience for practical object tracking problems using a fine-tuned object detection model with a centroid tracking algorithm works well enough. Much more computationally efficient and
154.
by alexcnwy 7y ago
Very cool and super inspiring, well done! Can you give any detail on how you're getting customers... you mention learning google campaign management, is all your traffic via google or have you found any other distribution channels that
155.
by alexcnwy 7y ago
I’m building a tool to speed up data entry. It highlights OCR‘d words with low confidence for human review. It uses templates for capturing structured data and a fast interface for capturing unstructured data. It grew out of consulting proj
156.
by alexcnwy 7y ago
> OCR doesn't need a cloud vs. a beefy computer - I do a lot of that too. Can you elaborate on this?
157.
by alexcnwy 7y ago
I use an outliner called Dynalist. It’s an alternative to Workflowy. Workflowy development stalled a while ago so I switched but it seems to be active again so I might switch back. An infinitely zoomable hierarchy works really well for my O
158.
by alexcnwy 7y ago
Excellent explanation! I find this visualization really helpful when I need to explain PCA: http://setosa.io/ev/principal-component-analysis/
159.
by alexcnwy 7y ago
1989/1990 was also when convolutional networks first started working with LeCun’s breakthrough paper on digit recognition. Incredible to think how much amazing research was happening back then and wonder what research is being done now
160.
by alexcnwy 7y ago
Very cool idea. It's not clear where you're using ML from the site - URLs seem chronological, not ranked on "relevance/importance" and I can't see any relevance/importance indicators. I'm curious to h
161.
by alexcnwy 7y ago
Great work - would love to read your thesis if it’s available online?
162.
by alexcnwy 7y ago
I did a 10 day retreat after burning out working at a hedge fund and I can highly highly recommend it. I must warn that it’s not easy. I thought a meditation retreat would be a great way to unplug and take it easy but the schedule is intens
163.
by alexcnwy 7y ago
Following the Twitter thread, keep up the great work Max!
164.
by alexcnwy 7y ago
Really interesting write-up! There was a great podcast about PPP on Darknet Diaries a few weeks ago - their relentlessness and persistence is so inspiring: https://darknetdiaries.com/episode/43/
165.
Numerai is giving away $1M in NMR in new Erasure grants program(medium.com)
5 points by alexcnwy 7y ago|0 comments
166.
by alexcnwy 8y ago
good question! I have wondered about this when plugging my phone in and being asked “do you trust this computer?”
167.
Deep Learning Indaba 2019 Kenya(deeplearningindaba.com)
5 points by alexcnwy 8y ago|0 comments
168.
by alexcnwy 8y ago
yeah it’s easy to fix those with a filter on the predictions. could use bayesian approach or just smooth using a majority vote over rolling window of say 3 frames...
169.
Thisartworkisalwaysonsale(thisartworkisalwaysonsale.com)
3 points by alexcnwy 8y ago|0 comments
170.
by alexcnwy 8y ago
experience earned IS a financial investment because it increases your future earning power. whether the increase in the expected present value of your future earnings justifies the present-day lower salary is a question of magnitudes and pr
171.
by alexcnwy 8y ago
so cool! i think this kind of thing getting more realistic is a backdoor way to get governments to pay attention to verifiable identity. what’s to stop someone from making a video of the CEO of a public company announcing bad news while sho
172.
by alexcnwy 8y ago
I think you’re right - it’s just surprising that 50 years after humankind put someone on the moon (allegedly lol) we still can’t get video conferencing right. I know it’s a hard problem but is it really that hard?
173.
by alexcnwy 8y ago
I understand that and perhaps it justifies the valuation but it blows my mind that in 2019 there isn’t an open source solution that works just as well (or better - again, Zoom is buggy) and is easy to deploy and control yoursef. I know the
174.
by alexcnwy 8y ago
I haven’t but I recently used their video conferencing software on calls with 40+ people and it was really buggy. All of those things sound great and important but there are many competitors who also do those and again, the switching costs
175.
by alexcnwy 8y ago
what do you mean?
176.
by alexcnwy 8y ago
blows my mind that video conferencing software with such low switching costs can be worth so much.
177.
by alexcnwy 8y ago
That only works with a small number of transmitters. If Samsung or Apple released a phone with native satellite default support, it would be impossible to police. Censorship is doomed.
178.
by alexcnwy 8y ago
when everyone has one
179.
by alexcnwy 8y ago
how do you imagine they would police it?
180.
by alexcnwy 8y ago
the internet will become impossible to censor (for everyone - it’s already easy for the rich to avoid censorship) when satellite connections become cheaper / faster / more accessible

More ›