What did you do in your shop today?

Lately things have been pretty weird. I got a Jetson Nano a few weeks ago and started playing with it. Determining location and surroundings via camera and lidar sensors seemed fascinating.

I fell down the computer vision rabbit hole. I spent more time than I care to admit trying to figure out enough linear algebra that I could understand enough of the math to even read about computer vision:) If that last sentence runs on and is a bit confusing... it correctly reflects how I feel about the subject.

I am not sure if I should continue to push through or just give up and accept the computer vision is a very cool black box that just works if configured correctly.

twobits,

I shouldn't worry about it too much, in the ever-present battle between Artificial Intelligence and Natural Stupidity, Natural Stupidity is still winning.

I was looking at Neural Networks using Python a few years ago, trying to figure out whether I could pick out winners or places in horse races. More useful than recognizing moving people or cars. I think!

@twobits

I used to get into the extreme details of how things worked when I was younger.  But now technology is growing at such a rapid pace that, unless you plan on making a career out of it, the black-box approach to just use the technology for its final result is probably the only reasonable way to go.

I don't yet have a Jetson Nano yet, and it's going to be quite a while before I get one. But my hope is that when I do get one I can just use it for the features its supposed to be able to do without having to learn how it actually does it.   Like for example, if its supposed to do face recognition, then I expect it to be able to do that without me having to learn linear algebra.

I'm also curious of just how sophisticated it is?   Does it just recognize the presence of a face?  Or is it sophisticated enough to actually recognize a specific person?   I'm kind of hoping for the latter.  If all it can do is recognize the presence of a face, but not be able to tell people apart, then I'm not sure if I would even be interested in getting one.   And I certainly don't want to have to learn linear algebra in order to be able to use it.

As you say:

I am not sure if I should continue to push through or just give up and accept the computer vision is a very cool black box that just works if configured correctly.

Unless it's your goal to lean how it actually works, then just using it as a black box.  That's what I'm hoping to do if I ever get one.

Yes, I agree life is too short to be spent learning math:)

The challenge is that the current generation of documentation seem to be written by practitioners for other practitioners. It assumes that the reader has a lot more prior knowledge than I do.

At this point, I was able to set up the Nano and run some of the example software. I have a bit of a programming background, but I am at a complete loss on how to incorporate the examples into something I would like them to do.

One downside of the Nano is the proprietary nature of CUDA. Many of the prolific hardware hackers and writers tend to avoid the proprietary stuff.

Thats a nice addition to the MeArm Robo. I have one out together and had it working with a joy shield, but the arm did not make it through the move to North Carolina 🙁 I need to see if I can print the parts that broke.

Anyway, I have been working kind of all over the place this year.

First, I decided to make some PCBs for various things - The Nano, Pro Mini, ESP Nodes, Z80, etc. I wanted to test the PCB manufacturing waters and did a few I needed/wanted.

Since I had new boards to play with, I started to build a Cosmic Elf type computer. Cheating a good bit 🙂

Also working on my bot.

Scott

Yes, that's another thing we need to watch out for with new technologies.  Sometimes we buy a technology only to discover that we're then stuck with having to buy more products or software from the same company in order to actually use the thing.

All this technology sounds interesting, but I don't want to allow it to get me off track.   My original goal was to build a robot.  And so I'm going to try to stick to that goal.  And currently my robot isn't at a point where it could use a Jetson Nano anyway.

One thing that I've discovered is that technology is becoming so cheap there is a temptation to buy cutting edge technology simply because it's affordable, when in reality we really don't even have a good use for it.

Not only this, but what's "Cutting Edge" today will basically be "Obsolete" a year from today.   Technology is advancing at such an extreme rate that nothing is going to stand still.

You could put a lot of time and effort into learning how to program a Jetson Nano only to discover next year that some new technology has replaced it.   I'm getting too old to try to keep up with such a rapidly changing world. ? 

Posted by: Robo Pi

One thing that I've discovered is that technology is becoming so cheap there is a temptation to buy cutting edge technology simply because it's affordable, when in reality we really don't even have a good use for it.

 

This is true. I have a "few" Orange Pi's because of this. 

Posted by: jscottbee

First, I decided to make some PCBs for various things - The Nano, Pro Mini, ESP Nodes, Z80, etc. I wanted to test the PCB manufacturing waters and did a few I needed/wanted.

That's cool.  Did you make those PCBs?   Also, I see  you mentioned the Z80.  I used to love that CPU.  I learned machine language programing on it using hex code.  I actually started writing a program with "Pop-up Windows".  And that was before Windows had even been invented. ? 

Having an OS that presents everything in the form of dedicated Windows seemed so natural.  It's hard for me to even believe that Bill Gates was able to copyright the idea.   But he did, and now the rest of us have to pay him for the idea.

As the proud owner of a Jetson Nano, I would have to suggest that it is not ready for use as the main compute unit in DB1.  It would be a cool enhancement to DB2. But there is a pretty steep learning curve for the interesting bits that make it 'better' than an RPI.

Posted by: Robo Pi

Posted by: jscottbee

First, I decided to make some PCBs for various things - The Nano, Pro Mini, ESP Nodes, Z80, etc. I wanted to test the PCB manufacturing waters and did a few I needed/wanted.

That's cool.  Did you make those PCBs?   Also, I see  you mentioned the Z80.  I used to love that CPU.  I learned machine language programing on it using hex code.  I actually started writing a program with "Pop-up Windows".  And that was before Windows had even been invented. ? 

Having an OS that presents everything in the form of dedicated Windows seemed so natural.  It's hard for me to even believe that Bill Gates was able to copyright the idea.   But he did, and now the rest of us have to pay him for the idea.

I laid them out in *cough* Fritzing and generated gerber files to send to a PCB house in China. JLCPCB. Good quality and super affordable.

I started by making PCB breadboard's that I liked and thought was affordable. Then I did a Nano and Pro mini board that allows for secondary power, to run servos and other things without bodging or taking the GPIO sourcing limits.  

Then I di the Z-80 board and others. I kind of lost my mind for a few weeks making them 🙂

 I plan to take some time and learn another PCB cad but am mixed on what. I have tried Eagle, Kicad, EasyPCB, and upverter. The jury is still out.

Scott

Posted by: twobits

But there is a pretty steep learning curve for the interesting bits that make it 'better' than an RPI.

That's what we have Bill for.   He climbs those steep learning curves for us so we can take the express elevator straight to the penthouse. ? 

Posted by: jscottbee

 

I plan to take some time and learn another PCB cad but am mixed on what. I have tried Eagle, Kicad, EasyPCB, and upverter. The jury is still out.

I'm still using a pretty old PCB program called FreePCB  It does what I need.  I also use TinyCad for the schematics.   I'm sure there are better programs available, but I have already created my own component libraries and I have designed them so that my components in TinyCad match up with my components in FreePCB.   So to move over to another software package would cause me to lose all that work.  I'm happy with TinyCad and FreePCB.  I typically don't work with large boards anyway.   So these old applications serve my needs.

Posted by: Robo Pi

Posted by: jscottbee

 

I plan to take some time and learn another PCB cad but am mixed on what. I have tried Eagle, Kicad, EasyPCB, and upverter. The jury is still out.

I'm still using a pretty old PCB program called FreePCB  It does what I need.  I also use TinyCad for the schematics.   I'm sure there are better programs available, but I have already created my own component libraries and I have designed them so that my components in TinyCad match up with my components in FreePCB.   So to move over to another software package would cause me to lose all that work.  I'm happy with TinyCad and FreePCB.  I typically don't work with large boards anyway.   So these old applications serve my needs.

I wonder if FreePCB was what I used in the early '90s to layout some boards I had done for work. I remember it was shareware (or maybe free) and I got it off a BBS in 1991-92.

Scott

Posted by: jscottbee

I wonder if FreePCB was what I used in the early '90s to layout some boards I had done for work. I remember it was shareware (or maybe free) and I got it off a BBS in 1991-92.

Scott

It probably was.  I'm pretty sure this was an old DOS program that was later modified to take advantage of Windows.