View on GitHub

Quorten Blog 1

First blog for all Quorten's blog-like writings

Red Hat notes. RHEL notes.

Useful information on getting the available versions of a particular package:

20170213/http://unix.stackexchange.com/questions/151689/how-can-i-instruct-yum-to-install-a-specific-version-of-package-x#151690

Important! If you can’t upgrade your RHEL 7.2 to the latest version of its packages due to some artificial hold-backs, what can you do? Here’s how to install a specific version of Docker that works on my older systems:

yum -y install docker-engine-1.12.2-6.el7.centos \
  docker-engine-selinux-1.12.6-1.el7.centos

Regardless, one thing that seems to stay consistent. Before you can successfully run machine learning algorithms, you’ve got to collect data and store it in some sort of standard file format. Yes, imaging file formats. So this is where more work, and, well, more buy-in needs to be done.

Okay, cool! That means that I would need to proceed by trying to build the… the what? The mass scanning project? The home inventory system? Oh, no, I know! The 3D scanning file format. 3D spatial information system, to gain insight into 3D data, to visualize it.


Crazy or not? Okay, now I found out not. Too lazy to scan the barcodes on objects? Well then you can have machines scan them for you with 3D scanning. Just modify the algorithm to cover the surface topology of the object, or operate across several 2D frames that give full object coverage.

So, this is when people say that even scanning barcodes is too much.

So, not crazy.

Of course, the ongoing problem with generic scanner data is that you must transform data from the sensor into some sort of format that can be presented on a digital display or printed. Often times the most useful transformation form is to make a duplicate of the object that is as close to the original as possible. In the case of sensor data that is meant for this target, the solution to the problem of presenting the data is rather simple. Just run through the process of making it into a duplicate of the real object.


So really, what separates versions of the same file format from different file formats? When you think about it, the main thing that separates here is the existence of forwards and backwards compatibility.

Oh, come on. Do I really need to use that L293D controller? I’d hate to add an extra chip to my setup. There’s got to be an easier way to do it with CMOS, isn’t there? Well, there would be, if only your GPIO pins had the option to be “ground” in addition to “0v” and “3.3v”, but they don’t.

UPDATE 2019-11-20: They do, but there is a limit on the voltage and current they can handle. Use level shifters, or a cheaper equivalent such as MOSFETs or buffer circuits.

20170129/https://en.wikipedia.org/wiki/General-purpose_input/output
20170129/https://en.wikipedia.org/wiki/CMOS
20170129/https://en.wikipedia.org/wiki/Gate_equivalent
20170129/https://en.wikipedia.org/wiki/Beyond_CMOS

Advancements in CMOS circuitry? Oh, this reminds me. I should go read IEEE Spectrum to update myself on technology news.

20170129/https://en.wikipedia.org/wiki/10_nanometer
20170129/https://en.wikipedia.org/wiki/7_nanometer

Oh yeah! Don’t forget. So, an important observation on 3D objects. I’ve noticed that many 3D objects in the manufactured world are very much the same in physical structure, differing only in color and surface printing. Yeah, this makes sense given the limitations of current manufacturing technology. But what else does it mean? It also means that the process of digitizing a 3D object is much simpler. In today’s world, all one needs to do is 3D scan a control group of 3D objects, and then 2D photographs of additional objects can be automatically matched against their 3D counterparts, and the remaining texture can be applied by simple matching means. Done, mass 3D problem solved.

But of course, it also begs a question to think for the future. In the future, with 3D printing, will we perhaps have a larger diversity in the world of industrial design? Could it be the case that every manufactured object is slightly different in 3D design detail from every other? Well, I don’t know, this may be an interesting future to come.

Okay, so I was looking around at the Raspberry camera stuff, and I found some interesting new things. In particular, a lot of useful 3D information resources.

20170129/https://tinkererblog.wordpress.com/2016/03/08/raspberry-pi-noir-with-ir-cut-switch/
20170129/https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=96997&p=673996#p673996
20170129/https://tinkererblog.wordpress.com/2015/07/28/how-i-designed-a-compact-weatherproof-raspberry-pi-case/

  • Particularly interesting on the article above was the number of times people talked about purchasing and payment requests to design a waterproof case for Raspberry Pi 3. Wow, very interesting.

20170129/https://tinkererblog.wordpress.com/2015/07/22/raspberry-pi-camera-face-recognition-and-email-alert/

Oh, wow, GrabCAD, another online repository of CAD designs, in addition to Thingiverse. It hosts a lot of open-source designs. But is it also open-source itself? Well, that I have yet to figure out, but I suspect that it is not as open-source as I’d like it to be.

20170129/https://en.wikipedia.org/wiki/GrabCAD

SketchUp 3D Warehouse? Max model size of 50 MB? It must not be that good, then.

Read on →

Whoa… so I’ve been fooled. You can turn a mobile phone into a video projector with ease. You only need a single magnifying glass lens.

20170117/http://content.photojojo.com/diy/turn-your-phone-into-a-photo-projector-for-1/

  • So that’s another use of the 4K smartphone screens, in addition to Google Cardboard VR.

Note that the article has links to more sophisticated projects too.

Gosh, now I have to reconsider… if you can do this many things with a smartphone, why do you even need the bigger computing devices?

Okay, never mind, no need to search for “orthographic projectors”… yeah, they do exist, just run an orthographic camera in reverse. But they are not quite as common.

  • You don’t need to wait for 4K projectors to come on the market if you use a 4K smartphone as your 4K projector. You might even have luck making an 8K projection screen if you put together four 4K smartphones.

Read on →

What about 3D reconstruction? You say it should work from just regular optical images, provided that you have enough of them? That’s exactly the issue! How can you entice a casual photographer to take that many photos of an object? Heck, they’re just too lazy! That’s precisely the issue. And that’s why it doesn’t work.

BUT, if you by chance happen to get lucky and there is some popular piece of architecture that many people come by to lazily photograph, then you’ll find that such techniques to be successful. In this case, it is a statement of reporting bias, that only the popular structures get scanned into 3D.

And, matter of fact, if such a structure really is that popular, chances are the geeks who have the equipment and determination would likewise come by and do a dedicated scan. So there you go. It’s kind of self-defeating, it is. Let me put it this way. 3D scanning objects is like taking pictures with a DSLR camera. It’s just one of those things that will never really become popular among casual, non-technical, non-professional users.

Oh, good idea. Web converter can also be used on local files too. In case someone doesn’t want to setup the command-line converter and wants compact storage formats for local files.

In the real world, you can typically only deflate hollow objects, but once you scan only the surface in 3D into the computer, you can deflate even solid objects. All you do is specify the outside and state that the interior is “solid.”

Read on →