As an aside: owls being wise is a cultural belief. In India, owls are considered unintelligent, and calling someone an owl is akin to calling them a fool. Interesting how these perceptions are diametrically opposite!
I too was inspired to use this dataset in some way, and made this toy program where you point your camera at your face, and it renders your eyes and nose as doodles (and barfs out more if you open your mouth): https://github.com/goberoi/sketch_face
Monkey Island and its sequel are some of the defining games of my childhood. My brother and I played all the Sierra Games, and Lucas Arts games, and while we have many loves (Heroes Quest/Quest for Glory, Indian Jones), Monkey Island was by far our favorite.
We few years ago we were lucky enough to meet Ron at PAX when he was there promoting Thimbleweed Park. Since then, I saw him fairly often at a neighborhood coffee shop here in Seattle.
Earlier this year, at the start of the pandemic, my brother and his partner ended up making an adventure game in the style of Monkey Island. It's almost an homage and has a few shout outs to those in the know. Check it out here:
Yes, this sort of tech is likely to make its way into CGI. Some relevant resources if you're interested:
1. This post shows how GANs were used to replicate the Princess Leia scene in Rogue One. The scene in the movie required a team of special effects artists, whereas this required access to lots of images of Carrie Fischer, a GPU, and this technology: https://io9.gizmodo.com/this-video-uses-the-power-of-deepfak...
tl;dr: Instagram is significantly tightening up access to their platform. The existing API is being deprecated in favor of a heavier permissions based system, and functionality available to all today, will be restricted to medium-to-large businesses down the road.
This is relevant to the HN community: builders who have enjoyed access to Instagram will now be severely restricted, which will limit future applications.
Finally: this is probably not a huge surprise. LinkedIn did this a couple of years ago, and of course Facebook has been tightening up access for years.
Why is this interesting? In short: a great new activation function that may challenge the dominance of ReLU.
Longer story:
Today, ReLU is the most popular activation function for deep networks (along with its variants like leaky ReLU or parametric ReLU).
This paper from the Google Brain team is ~2 weeks old, and shows that SWISH, a new activation function, "improves top-1 classification accuracy on ImageNet by 0.9% for Mobile NASNetA
and 0.6% for Inception-ResNet-v2" by simply replacing ReLU with SWISH.
SWISH is equal to x * sigmoid(x), so not that much harder to compute either.
They are going to expose some of this functionality as part of their AR libraries. We know this from their Snapchat demo, as well as the gaming demo (though that used the rear camera, not front one).
So... yes, developers will be able to get their hands on some of this tech, but how much can we do with it?
If room mapping was truly possible, I would imagine they would have done more in their demo with this, e.g. see the Hololens game where robots come flying out of walls, and hide behind your couch. That demo was more substantial than seeing a flat tabletop with a projected 3d game, or a projected 3d robots standing on a flat basketball court.
Can the new camera technology be used to map spaces as big as rooms?
If so, this could be game changing for allowing the creation of VR spaces quickly and inexpensively. E.g., play a VR game in your real house after mapping it with your iPhone X. Or better, do a detailed remodel in VR before doing it in real life.
A lot of commercial uses of VR technology (e.g., construction, industrial design, etc.) can benefit from inexpensive, and accurate mapping. Today, the alternative is to get an architect to build a model of your house, to build a crude version yourself, have a Hololens/Tango phone/or other nascent and expensive technology.
If not, what truly is the game changing aspect of these cameras + specialized compute for machine learning/neural nets? They have to have thought through dozens of use cases beyond photos, animated emojis, and other trivial entertainment... right?
Did I miss it, or are there no virtual reality startups on this list?
I'm a wee bit surprised: VR is nascent and there aren't enough devices out there to build sustainable businesses shipping content, but I would have expected at least 1 or 2.
Yeah, Clarifai did well. I am keen to learn how well their custom model feature works. Per their FAQ[0], you only need supply 20-50 images per concept. That seems remarkable to me, given that a concept like 'cow' has ~1500 images on Imagenet[1]. Perhaps they are using some sort of transfer learning to facilitate this? I.e. using a pretrained model, and then only retraining the last few fully connected layers, or retraining parts of the entire network?
I am not a deep learning practitioner, but would be curious to know from experts how their custom model feature might work; and from any of their users on how well it actually does.
Tablas: haha, great description of your teacher. I do play, with enthusiasm, but poorly. For those in Seattle, there is an amazing teacher who teaches up on Cap Hill [2].
These aren't really great for learning about computer vision or deep learning, but are great for building projects that require image classification.
E.g., recall the awesome project for lego sorting by jacquesm[0] ? He built his own model using Keras and Tensorflow, but you may be able to achieve similar results by using Clarifai's feature to train your own models with no understanding of deep learning. This is great if your goal is to build a thing, like a lego sorter, but not so much if you want to learn how to build a state of the art image classifier.
If you're interested in learning about computer vision or deep learning, I recommend searching this site to find threads that cover that extensively. Good luck!
At the time, Clarifai was the "best" one (I caveat by quoting because this was a for a small corpus, with subjective results, not a real train-test cycle). I re-ran the results about a month ago (linked to from the post), and found that Google and others have continued to invest and improve.
I haven't built a production app, but did compare the computer vision offering in Microsoft's Cognitive Services with those offered by Google, IBM, Clarifai, and a couple of others.
Summary (taken from the post): "In my small corpus, I would rate Clarifai at the top, followed by Google and Microsoft at approximately the same quality, and then well below them IBM."
If you're interested in the genre, "Hunt the Yeti" is a new adventure game that was released a few days ago and seems to be well received based on the reviews (I have not yet played it).
There are lots of apps that connect to established third parties that Amazon is likely not to build, e.g., home automation with SmartThings, music with Pandora. And also several silly but fun things that they may not wish to maintain, e.g., jeopardy game or the fart app.
But yeah, I don't blame you for being cautious of their intentions. Developers have been burned in the past on these big platforms (e.g., Twitter [1]), but I have hope that they will realize the pie is bigger if they make their platform attractive to 3rd party devs.