Everybody ale.

Welcome to a another part of the python plays Grand Theft Auto series.

In this video, what I'd be doing is attempting to put tents or flows Object detection, a p I on to this machine.

So currently, I am just upgraded tensorflow.

So we should be on import tensor, uh, tensor flow as t f.

And the conversion should be 1.11 Release candidate zero.

Now, I'm a little concerned that this may or may not work very well, because the object detection FBI historically has been quite finicky.

Especially when it comes to the, you know, pro Taki, all that stuff.

So And I think tensorflow one after, like, 1.9 or something like that requires a specific version.

Proto buff, And then you're going to need a specific version to work with object detection.

So that'll be fun.

All right, Eh?

So let's go ahead, get started.

I'm gonna be using a few things.

First of all, python purring that net the, um the object detection ap i tutorial I did should prove to be somewhat useful in this endeavor, and then also, if you just google, um Let's do tensor flow.

Object detection.

A p I look for videos.

This one right here from EJ Electron ICS.

Um hey, it's, uh I'm gonna hop to the get hub here, and, uh, we're going to make use of a few the commands here, as I went through.

So I looked to see if I could find any more recent than mine tutorial on doing 10 frivolous object detection.

A p I and I found this one.

And at least this one is updated to tensorflow 1.7.

Um, and then the python path stuff apparently is not even required.

Anyway, the video is for 1.5, and I noticed there are a few different commands.

Also, the pro tac command has changed pretty significantly, so I'm pretty sure those two things are gonna be a problem.

Plus, in the comments section I was seeing people were saying, like, this is the wild card.

It doesn't work anymore.

Which is what a weird thing to change this like, it's like sloshing around.

Hopefully y'all didn't hear that, but you probably did.

Okay, So, uh, so that's what we're gonna be working on.

So, first of all, well, the problem.

More important thing is tensor flow object detection.

Ap I Let's go.

The official docks here.

Maybe I'll scroll.

I can see this.

Really?

Well, it's just This is probably pretty small for you guys.

But anyway, uh, scrolling on down to insulation, all right, Because you don't just need this instant.

This is the truth.

From the very, very beginning.

It's like you need more than this installation to get this thing working for real.

Like, I wish you could get the full installation of the tensorflow object.

That section e p I using their install suggestions.

But you can't.

Okay, so we've already got tensorflow GPU have already?

Well, I haven't shown you, but TF version.

We may end up needing to change this.

It's It's a requirement for me that we are able to run tensorflow object detection.

So, um, so I may end up downgrading this.

I really don't want to write my own tensorflow object detection.

A p I I mean, there's a lot of work going in and this is actually a really great a p I, but it could just be tedious to use and set up, but ah, the first thing that we're gonna go ahead and do is install, You know, all these little requirements here.

So I'm just going to also wonder if I can, uh, dang it.

I see, uh, properties find Let's go with a good old 24.

Okay, let's quit out of here, and we're gonna do make sure Pip is Python 36 it is.

All right.

So we need to been stole all of the society thon context living on that.

So I'm gonna run Pip Install.

Uh, I think we should be able We should be safe to upgrade all of these.

So Cy Thon?

No common man context.

Lived to pillow l xml, Jupiter Jew bidder and Matt plot lib.

Okay, so this is, like there.

Let's see.

I think they do differentiate usually between windows in Lennox or I thought they did.

I don't see it here.

Cool.

Thanks, guys.

Um, but one of the other things we need is, um, that prote above stuff.

So next thing I'm gonna do is bebop on over.

Oh, maybe it was me that differentiated between windows anyway.

Yeah, they don't even bother telling you like Hey, by the way, on windows you're gonna have to take some extra steps.

Okay.

So, um, so we're gonna head over to the pro talk release page.

Care us, but you'll have care is incompatible with the show.

Must go on.

Serves.

Wait.

We didn't upgrade.

Hopefully it doesn't break.

What?

Hold on.

Go back.

Huh?

Ah, where is the Okay?

Well, we'll deal with that in a minute.

So first of all, it suggests that I think we can get away with three plus pro talk.

Um, from what I could see in the comments section, people were just saying that you can't use the wild card, which isn't a big deal.

So I'm going to because this lovely, lovely person ej electron ICS um, it was somewhere in section two.

Somewhere somewhere, it's like a really rather than using the wild card.

He just wrote out all a love them s Oh, good work, soldier.

Um, we're gonna make use of that, so I'm gonna try to get away with 36 I don't even know.

Let's see if this guy mentions Uh oh, my gosh.

This scroll wheel on here needs to go faster.

Let me just do a pro talk.

So should be before the compile pro tox.

When did he proto portable files have not been compiled.

Oh, he uses conduct to install proto buff.

I don't think we can do that.

Um, hip install pro buff, pro talk.

Yeah, If somebody knows how we can use protoblood if I would love to know, like, just by, like, a pip install But I don't think that's ever gonna work out for you unless maybe we do the padding stuff, but I don't think that's gonna work anyway, So we've got the installs.

That's basically the first thing that we want to dio.

Next thing we want to do is go ahead and clone this stuff.

So I'm going to Not this one.

I want to clone.

Oh, Well, while we're here, let's grab pro talk.

We might have to downgrade pro talk, but for now, I'm just gonna grab it here.

Protect 361 We might need to go with, like, a 3.4 or even a 3.0, the docks specifically say three point.

Oh, but I know 3.4 works because my latest tutorial did 3.4.

So, um, try silent stuff this I think I'm gonna throw it in.

See, though, to be honest.

And where did you hear is?

So I'm going to do we actually later.

Let's create a new directory.

I'm gonna call it pro talk.

It's may or may not be a great idea, but I'm gonna do it anyways and extract all cool.

So now, pro.

Uh gross.

What did it do that?

That's not what I wanted.

Hold on.

Opening new window cut pastes and I'm gonna remove these.

Okay, so now when we want a reference pro talk, we just specify the path C colon slash pro talk.

Easy enough.

So, uh, now what we want to do is cologne, this entire, like models directory from Tensorflow s.

So I'm just gonna click on this mean here, So models, and then it has, like, slash research and stuff, which was kind of a hardship for people thio overcome.

At some point, they, like, added the research directory, I guess for, um, organizations sake.

So go ahead and download.

Uh, the models folder.

The shark makes him some great noises, and I think I think I'm gonna throw it to the desktop for now.

Trying to decide if that's what really where it wanted.

Um, I guess I'll leave it in the desktop.

So anyways, it's on the desktop.

Extract all.

Oh, my gosh.

Please, no.

Is it really gonna take five minutes to extract this?

Ah, gross, man.

Dang it.

We should have been doing this while we're doing other stuff.

Then get Okay, So let's see, What else could we do while we wait on the okay?

I think it's gonna speed up for us.

So I guess, you know, we just need to wait.

So I guess I'll pause now and then.

We can continue on when this is when this is done.

Although you guys are all gonna have to sit through it anyways.

Uh huh.

Maybe you guys can extract faster than me.

This feels like it's, like, super super slow.

Um, so once we have this all extracted, the next thing we're gonna do is use the protocol buffer to compile all these things.

So going overto edge electronics, I'll try to remember the bottling, the description.

But if I forget which I brought, it might just google attentional object detection.

A p i you should find this.

You should recognize this photo?

His car on his head.

Uh, anyway, um, and then you should find your way to this.

Get hub.

At least that's what I'm using for Now, we'll see if this works.

And I'm sure we'll have to change some stuff.

Although it looks like maybe he's been updating it over time, which is nice.

S O c models.

Where did it go?

Here.

It's hard to see that on this desktop.

Okay, so, models, master, and then this is it.

So the first thing we'll do is we're gonna rename this system models.

It's kind of silly to be models, master.

And then this, um I guess I'll leave that for some reason, this extracts differently.

Then I think my home computer does, like normally if I extracted models, master, it would extract just the contents and not start with this original directory anyway.

So the next thing we want to do is go down.

Thio.

I think for him it was like step two f right.

And I think the next thing we want to do here is run these things pro tac command again.

I'm not really sure.

I guess he could do it because of Anaconda.

So if you have Anaconda, you could just do the condom install proto buff.

But this whole thing needs to be run.

So the first thing we're gonna do is just copy.

I'm gonna hit.

Copy.

I'm gonna open up like a note pad or something, because this needs to not be pro.

Talking is B C colon slash pro talk that should work and then python out.

And then all of these things And he's running this from object, that texture.

So object detection needs to be so research.

And then he's running object detection.

So Right, so hopefully from here we should be able get away with this.

So it's open up a command prompt.

Really pretty sure.

Oh.

Ah, do I need Ah, right.

Let's put some quotes around this, maybe.

Oh, you know what I'm doing wrong.

Um, so it should be pro talk, and then it's probably been and then pro talk, So c colon slash pro talk.

Been pro top pro talk.

Copy.

Okay, so we got all those.

Great.

Now the next thing, Um, I'm actually not sure.

I don't think in the past that was required.

The building install um I don't think it is because we're running from the directory.

Maybe if we wanted to use it outside of the models, let's let's find out.

So, research, let's go down to object detection.

Hello?

Hello.

There we are.

Um, and we should be it.

No, I did not click that.

Stop it.

Um, let's see what else he's doing here.

I think cause he's also doing, like, the creation of your automatic creating your own object detection.

I don't think we're gonna need to do that.

I just wanted at least here get to the point of running the actual things.

Okay, so we got the Jupiter notebook here.

So if I do this Jupiter notebook, let's see if this thing pops up well or not.

Cool.

Okay, so then we're gonna open up the object detection tutorial, and here is where we probably fail.

This will be the first failure point.