#2D Bounding Box

[hanszoe]

i think i'm too tired to work much on my image viewer build tonight but i'll try in the morning

[inksandpensblog]

The Box

I'm finally watching the episode! These are my stream-of-consciousness thoughts through my initial viewing. This isn't thought-out refined analysis, these are just the thoughts as I have them the moment they appear in my head.

They're getting better at the 3d running/walking animation.

Green being the one to notice details, like in The Village.

Civilization!

Green only drew one but then he just has three more? Cartoon magic or stickfigure magic? Or just saving time on a meta level.

This looks like a nice little toh that's a white-filled head, they're ALL white-filled heads (all look like freehanded circles though)

For a given value of "white," I guess, but they're all very light colors. And they all have head accessories. Interesting...

Okay it's kinda funny that nobody would talk to them and the truck is right there. Did they think the color quartet were giving handouts or something? None of them stopped to listen long enough XD

Ooh, so these are some kind of simulation chambers? Looks like they've simulated a desktop art program.

Hmm, so Orange's ability to just have whatever he wants manifest into existence with the stroke of a pencil is unique to him; these guys try the same motions and all they get are regular lines. (Also if Orange can do all that outside then why are they doing these experiments in the simulation chamber? Is it just for safety or can they not do it outside? Shady could use his toolbar outside just fine...)

Orange's drawings were alive, almost from the first line it seems. Before they even took their shape. These ones aren't.

I don't like how coldly they're treating my boy :(

you.

and then they don't even interact?? okay, so...Victim is pretty single-mindedly focused on whatever is about to happen with Chosen, and whatever it is doesn't involve Orange so Orange is irrelevant to him rn. Why bind him then? Where's he going? :(

he went in. by himself. I'm...not sure how to feel about that. surprised? it's not like the thing had a door. though looking at that control-display, it does look like there's something within the chamber itself that's also containing Chosen. Or is that just the bounding box? I think it's a bounding box actually. So this is also a simulation chamber. (this is giving me all sorts of questions about how the artboard recognizes the toolbar icons as graphics while the tools still retain their functions as tools and I'm gonna stop before I spiral)

oh, it did have a door

WAITAMINUTE why is Victim's bounding box 2d and Chosen's 3d???

...oh. so...Chosen's fire registers as a separate graphic from him, then. good to know??

wait what the- just altering the visual is enough to affect his ability in producing that power??

why did we just cut to the hunter sharpening their spear? don't like that. not while you're stealing all my boy's natural defenses.

that one lingering second after Chosen's lasers die out but before he turns around to face Victim is haunting. Bro's been trapped and depowered again.

Okay nitpick but Victim's head isn't changing shape every frame and it's driving me up the wall. Did he have some work done between the last installment and this one?

I can't tell if I'm supposed to laugh or not at Chosen just standing there and tanking all of Victim's hits, and honestly I like that I can't tell, the dissonance is fun. But putting that aside...Chosen just let the dude have at him like it was no big deal. He didn't actually fight back until he decided he'd had enough. I suppose he could've just been confused about why this guy was attacking him...but he doesn't even block or dodge. He doesn't defend himself, his hands aren't even up, he's not even really in a fighting stance, and it takes him a bit to even retaliate. Even after the first one, his body language is just "bruh why." Which. Is an interesting attitude to have when he's trapped and depowered and being punched by a stranger.

Chosen finally counters, and it cuts to the control panel outside, looks like they're about to raise some setting. I've gotta take a break for a meeting, here's my thought: Victim never left the art program, in AvA1. In a very literal sense, he spent his whole life in it. It's where he found all the things that enabled him to fight back. In a weird way, it makes sense that this is where he's powerful, that this is where he's comfortable fighting. (and I know my personal headcanons are about to get blown outta the water but I'm really glad that I was on the right track in regard to this idea. though something I hadn't ever imagined was that he'd rely on other people to supply those advantages instead of obtaining them himself)

Meeting over, back to the episode! (I wonder if the title of this episode refers to the white chamber Chosen is in or the actual bounding box. Given that the mere appearance of a bounding box in Wanted was enough for many of us to clue in to Victim's involvement, even before he appeared in-person at the end of the episode, I'm leaning that direction. Amazing that it only happened once in the series before Wanted and it was already enough of a legacy to make him recognizable. I'm now realizing that in AvA2 the first thing Chosen did even before attacking the cursor was destroy the bounding box. Also realizing that the bounding boxes in this chamber aren't visible, unlike in the test-chambers in the previous scenes.)

interesting...each of Vicitm's limbs are separate graphics with their own bounding boxes, which is how they looked in the Bloop animation course that Alan made, but it isn't how he was in AvA1. Though he was recognized as a single graphic just a few scenes ago, so that might not mean anything.

why, stop, dude we get it you're strong now you've proved your point stop beating him up-

oh good Chosen's still got his own strength, he's fighting ba- what- okay what, can't take what you dish out??

Ooh, I don't think we've seen Chosen actually throw fire like that (instead of just blasting and halting blasts) since the early episodes.

okay, nice to know that Victim actually can dodge on his own

...don't like how little time it took Chosen to start feeling exhaustion...

...don't like how much trepidation the lasso is giving me. that feels like one of the crueler things to use against Chosen; not because we've seen it used against him before (only Victim and Orange have used it themselves, and only Victim has had it used against him, and good grief Orange and Chosen still don't know how Victim is connected to Alan-) but just because...I really don't want him bound and leashed again, he's been through enough of that. still, I had predicted that we'd see him using the lasso at some point in AvA6. I'm not happy I was right. (man, back before we knew Victim would make a proper return to the series I used to love the idea of him using it again. it's like the clearest example of him outwitting the animator and taking his tools for his own use.)

NOT THE NECK- they've never gone there, I don't like this-

oh what they actually- I was not expecting that. whips haven't appeared in the series before. oh I really don't like this.

God he's crawling back- why-

I had to pause. either the group outside just did something with the controls that Victim wasn't anticipating, or...Chosen felt threatened enough to turn his powers on himself and encase himself in ice as a defense. fuck. someone get him out of there. no flight, can't fight, man was pushed far enough he literally resorted to freeze. The Chosen One. I don't think that's ever happened before.

Fuck, that's just too much, I can't even bring myself to feel excited about seeing the duplicates again. though it is good to know that my headcanon about him needing to be in an art program in order to duplicate himself ended up being accurate. we'll see if my headcanon about it being one Victim with five bodies (as opposed to being five Victims) holds up. if they even have a way to show any difference.

[sighs, pulls AvA1 up in another tab] yep, they're all there. lasso, hammer, extended thumbtack (though it doesn't look like a thumbtack here, which is interesting), whatever that chain-accordion thing is, and the ninja star. why are you doing this? man literally froze himself immobile to get you to stop torturing him, just leave him be.

...haHAHAHA! Oh that felt good, that made me smile. Hopefully this means Chosen has recuperated a bit.

the animation of Victim on fire was so nice I completely missed Chosen breaking the ice XD going back to catch that made me realize that the other Victim duplicates literally just...stood and watched the one burning...

don't let yourself get backed into a corner-

OH he escaped the hold, nice-

I didn't realize until the lasso came back that he'd gotten rid of it when he broke the ice :( but before that he actually uses his fire-breath again, which is always nice to see. interesting to see it concentrated in a thin stream instead of just roaring out like usual.

oh yikes why- him on fire-

...don't do it, man...

OH HE OWNED THAT, LET'S GO!!! Great thinking, Chosen! (heh, nothing about this situation remotely resembles that time Chosen grabbed the cursor and forced it to click him free, but I was reminded of it nonetheless. something about the tool being turned against the user, but in such a different way than Victim does it.)

OH HE- ...oh...I was gonna say oh he learned, he adapted enough that he predicted them flipping his fire and he used that- but then they just nullified it immediately...

Victim don't snap the rope like that, we don't need any more implications we already know you're planning to put him in a world of hurt, why can't you just leave him alone now, what are you trying to prove

hh, back to Orange I guess...

MATH SPOTTED, MATH SPOTTED-

oh! they're bringing him to the- okay that makes sense, if it only works when he does it then obviously the next step to figuring it out is gonna involve him.

wait was he- could he not move when the bounding box was picked up?? (also neat that he immediately got worried once he realized he's in a bounding box, dunno if that's because of what happened in the last episode or if it's just because...he spends a lot of time in an art program so he'd naturally be familiar with what they can do and what it means to be inside one)

...well, nice that they at least get his attention before picking him up

oh now you're being friendly, sure. yeah Orange ain't having it.

okay, [picks up pencil] [instant notetaking] was funny XD

aaw, it's swimming around him

oh it knows what the eraser means

...wait it wasn't an electric eel in the last episode-

hexagons- and the power flickered

okay Orange stumbling around trying to stay out of its way makes it clear that he isn't controlling this thing once he's finished drawing it, it's acting on its own. it...it really feels distressed.

hah, they're all cowering. losers.

oh, back to this.

...did you have to make it so personal, Victim?

a chair

...okay not sure why that happened, could Chosen not just...lean forward? I rewound to watch it again and realized he's exhausted again after the slow-mo ends (is that from the slow-mo itself or just his exertion from the fight before the slow-mo?) and he...doesn't even struggle when Victim lifts him by hand with a lasso to the neck. is...is he just ragdolling now? is he at the point where that's his best option for minimizing conflict?

...he is. he's ragdolling. fuck. (damn you cc!Alan for introducing ragdolling in a comedic short-)

...no? no what?? no, he doesn't know anything about the animator? (which we know isn't true, but why would he deny that?) or no, he isn't going to answer?

don't tie him up, please...

oh now he starts struggling?

...back to ragdolling. good grief, the way they animate his legs just swinging even though they can still touch the floor...

...heh, I appreciate his spirit, especially at this point, but...Chosen you kinda just put yourself in a worse position.

...oh, what is that? I don't think we've seen anything like that before. It's got the floppy disk that's used as the "save" icon in many applications, but it looks kinda like a headset.

...a VR headset specifically, I guess.

OH WHAT- okay rad animation, but WHAT

...oh fuck, I've always wondered if there was a way to see the data that would be a stick's memory. FUCK what are they gonna use this for

wait they're bring Orange here?! Or are they just passing on their way back to the cell? but why are they going back to the cell, was the eel too much for them??

...wait, what? signal lost, I assume because Chosen dislodged it enough to disconnect, and then...instead of picking up where it left off it jumps all the way back to Showdown?? was...was Showdown just on Chosen's mind? or is this the memory of him recalling it right before flying to alanspc to entreat Orange's aid? ...or...are memories from beyond the sky-barrier not compatible? that doesn't make sense...

what's with everyone's reactions to Second's powers?

...I forgot Orange's cell is in here. whoops.

oh this is how he finds out??

hands first ("I did that??") and then his eyes ("Chosen was right??")

...and of course. I expected they'd lock him down with extra security as soon as they realized, but. poor guy.

(dammit I purposefully refrained from making a "dark mode" joke, you didn't have to go and put it on the damn control screen)

...and even after that (which, judging by Victim's reaction, even he hadn't known about Second's powers), the animator is still Victim's priority.

oh, this isn't just any clip of the cursor's involvement in Showdown, this is specifically while Yellow was rendering the cursor in.

and then they immediately pan to Yellow which means that observation is actually gonna be relevant, FUCK-

OH FUCK THE WANTED POSTERS

("earnings growth," so they actually do have a business element to them)

aaand the security cameras caught them peeking out of the truck, they literally just started printing the posters and they've already located Yellow-

that fight is mesmerizing. we haven't seen one like this before, as far as direction and setting. plus it's really interesting to see how the four respond. I might look deeper into that later. also that was a cool transition.

and Yellow is dragged off. The story has never had one of the quartet singled out in-universe by the plot like this before. they aren't even apprehending the others.

...of all the sticks to have a lighter, I never thought it would be Victim.

to be continued. good grief.

I have to eat before work so I'll think on all this and say more later.

[wormheamer]

wormboard devlog 1/?

there are two main things i've been working on lately, which are

(a) reducing the amount of space brushstroke data takes up on disk : not a huge problem right now, but crosses over to saving on bandwidth when i get around to networking stuff.

and (b) partitioning the infinite canvas. up until now i've just been drawing the whole canvas every frame, which gets exponentially worse the more that gets drawn. it still performs ok, since this is a decent computer and i can only draw so many lines so quickly, but all the public whiteboard fox stuff i've taken part in have been SUPER slow.

problem A wasn't too bad! each brushstroke is made up of two things: one byte for a color value, 0-255, an 8-byte unsigned integer holding the number of points in the line, and then that number of point values. each point is three single precision floating-point numbers, 4 bytes each for x position, y position, and size.

we probably won't ever need more than 65535 points in a single stroke --- 8 bytes down to 2.

for points, three more things need to be stored: the largest point size (S), the longest distance between two points (D), and the absolute (x0, y0) coordinates of the first point. then instead of storing absolute coordinates, we have them be relative to those --- x and y are signed bytes scaled down to 1/127th intervals of D, then added to the position of the last point. size, likewise, is stored as 255ths of S.

so long as the drifting produced by this loss of precision is kept track of and corrected for, it all works out pretty well! in the absolute worst case, only a single point, a stroke goes from 21 bytes under the old model to 20 under the new.

not much to write home about…

but a more typical stroke, 20 points long, comes from 249 bytes down to 77. much better!

old_size(n) = 9 + 12n new_size(n) = 20 + 3(n-1)

as for the canvas partitioning…

it's not just for rendering performance that i wanna get it done, but also for erasing and selecting and … basically everything, really? it's pretty central.

my first attempt was, as they say, "Жил был король когда-то, При нем блоха жила." it stinks.

basically i tried to create an infinite grid of fixed-size squares as a nested linked list: the canvas has a list of rows, each row a list of columns, each column tile a list of pointers to brushstrokes. it … kinda worked. not super well, because my implementation was bugged to hell and back, but okayish. if fixed and polished it could be decent for small-to-medium sized canvases.

but for the public whiteboards i've been in, people will tend to go off on their own and find their own corners, or draw really big pictures of rivulet encircling the entire canvas, or whatever. and when you try to do that ---

very quickly the list grid becomes untenable. most of those tiles completely empty! they only exist to create a path to get to where stuff's been drawn. it's wasteful of memory, and it's confusing, and it's slow, because the program constantly has to scan through one tile at a time, bugs in a maze.

but i want a really infinite canvas, where empty space takes up no … space. i'm very funny i know.

i'm not gonna write about the solution to that today, because i don't know what it is yet. removed the first grid implementation in favor of just testing the bounding boxes of each stroke individually to see if it's on screen, which is performing way better by virtue of how bad attempt #1 was.

the basic problem --- how to iterate over the objects in a certain area, and only those, with as little processing as possible --- seems adjacent to collision detection, so i might have to read up on the data structures they use for 2d physics simulations and such.

it occurs to me just now that i …. didn't actually need to be creating so many in-between tiles? could just search up to the edge, create a new one only if there's a suitable gap. might revisit that later, but i want to try out something a bit smarter…

quadtrees seem interesting, as a way to divide finite space, but again i need something that can grow outwards dynamically. k-d trees are the most promising thing i've seen so far, though i'm not sure i understand quite how they work yet --- something like a two-dimensional binary search tree, where you search first on one axis and then the other? they're used a lot for searching through point clouds and such, which is similar to what i'm doing. a sparse grid of canvas tiles is just a special case of a point cloud, probably.

thanks for reading

[nikoisme]

Something that I always really like in your art is your use of shapes. They're simplified but still very believable if that makes sense. Do you have any tips on that? Or could you direct me to where you learnt that?

I'm glad you like the shapes! They're one of my favorite things to draw and see in art hahaha. I'm going to explain how I personally do them, so it might vary for others.

So when looking at objects, i do try to break them down in big shapes. I usually hold what i'm drawing as a 3d object in my mind, and then i just try to,,,, rotate it. It gives me a much better sense of how it looks. For example, if i'm drawing a head, i'm drawing it as a box - i can easily rotate a box in my head, so it gives me a stronger idea of how i want to draw it. Also i just try to feel something out. I can feel out the structure of my finger or face just by touching it. This helps me a lot for planes, which comes in handy when i try to render things.

Instead of looking at objects as a bunch of different, smaller lines, i try to break them down into bigger lines that kind of define a shape. For example, if i'm drawing a face, i'm not going to draw each individual slope or curve of a line. While it would look fine, i would have no idea what i'm trying to draw. I'm drawing lines, which are 2d, instead of shapes, which are 3d.

Here's a very rough example

(the angles are a bit different but you get the idea)

With the first one i tried to draw each little curve of the face, but in the other i just broke it down into big lines/shapes. While both look okay, the second one is much quicker to do and a bit more shaped. I want each line to count. So i have one line for the cheekbone, one for the jaw, etc. Also, most of my sketches are just a bunch of boxes, so i have shapes as a fundemental. Of course you can use any other shape, but i personally use boxes because they're the easiest for me to work with.

Then the one that definitely helped out the most was doing a shit ton of figure studies. Instead of focusing on the muscles and the details, my goal is to get the overall form down. For some reason i am horrible at drawing in large scale, so the figures are really small. And I just hold the fact that i want each line to count, breaking things into big shapes/lines and just rotating things in 3d in mind as i go.

These are super rough, and i can definitely spot some mistakes - but the idea is there. Almost each big muscle is a single line. If i have a good sense of the fundamentals, then i can build up on the rest of it.

What helps a lot is of course, practice. I have pages upon pages upon pages of just random poses and heads and other objects in slightly different angles. So something is bound to stick. Also getting a good sense of the fundamentals, aka anatomy. Skulls, muscles, facial planes, bones, you know the stuff. If you understand the fundamental rules, you can break them and adjust them however you like because you know how things are supposed to connect to make sense. Sinix on youtube has some really really great videos on the fundamentals of anatomy and how you can break things down into shapes! Also you can find those 3d references where you can rotate the thing you want to draw.

Also just do whatever shapes you like best! I personally prefer boxes (or squares and rectangled for that matter), i find they suit what i want best. The heads i draw are boxes, the fingers, the hands, the eyes - hell i will break a round object into boxes if i so desire.

Most of this is internalized at this point and became muscle memory, and honestly i never really paid it much attention. So apologies if this doesn't make much sense hahaha! Shapes in general are something i love seeing in art, so i probably absorbed some other things through osmosis (aka i just,, stare at things a lot. I will see a fucking lamppost and go "wow those are some really good shapes you have there" and in my mind i'm already breaking it down like "this is a cylinder, one big line for this side, a box there, some slopes and curves on the other side" so i can reference it from memory later). I mostly fouced on the human figure here, as that's something i draw the most - but this counts for literally any object! I could do figure studies of a chair, why not. My process is the same.

[catless-and-covenless]

Xxx(Authors Note: Towards the end of the drive, while Gil is looking for parking)xxX

I guess I just wished I could immediately believe her. I wish I could full heartedly believe this change. There’s a very small part of me that’s hesitant and pessimistic that this is only a short lived change brought on by whatever they saw in the house. Or a change brought on in front of the others. So obviously it's time to word vomit so I can purge it on here

I worry. I want to not think this way. I just….I don’t want to be envied. I don’t want to be this ideal. I guess it still feels like…like a different type of objectification. For people to think I’m perfect. I’m not. And there’s something frustrating and-i don’t know-scary-? About that. I hope that’s not how everyone in the MBoM sees me. Because there’s a part of me that just wants to break down and cry. I need to. I can feel the purple box getting close to overfill. I need to go home and make myself cry by watching movies.

But frustrating because it feels like I can't, I can't because I have to make sure no one is fighting. That people make up before the tectonic plates either cause a volcanic eruption or an earthquake or tsunami. It’s frustrating to hear that me doing my best to help others try and be their best was perceived and received wrong. Like no matter what I do I’m wrong. Like I can either cause fear and hate or envy and annoyance. If I try to be the bigger person that just adds to the feelings of me being perfect and it’s frustrating to know that I can’t just be nice to be nice. I must be “being perfect” by backing down or flirting. It’s frustrating that I’m not better. That even with my friends I default to flirting when I get nervous or unsettled. It’s frustrating that’s one of the only times I am listened to. It's frustrating I can't make myself be loud and angry. I just don't want to be seen as a wolf either. It's frustrating because by being seen as perfect-No wonder Cat would turn on me. Perfect people aren't allowed mistakes.

And it’s scary. If someone sees me like this, what's to say the others don’t? That I’m either this 2d evil witch who hurt R or this perfect all knowing witch. I’m not perfect. They don’t know my throat is still aching from the time I bound that stupid lizard. That I don’t know all the mythology I can. That I’m limited. That It is dependent on my connection to the Abyss. A connection I only started feeding regularly. A connection that’s still new and I don’t want to abuse. A connection I want to nurture. What if my connection is weak or my concentration shot or I'm tired or sleepy. What if when they need me I won't be able to provide. I've known and played in the Abyss longer than most of them. Believed these things and know more than most. But I am still just human. What if I fail and I'm suddenly not perfect and everything becomes my fault again. What if I find my weak spot in front of everyone.

I just want to be seen as Xochitl. Not perfect or amazing or or sexy any of those things. I just wish it was enough to be me. But I guess that's the life of brujas. We live to serve. The first to offer help and the first to be blamed. Something something with great power comes great responsibility. And I do know more. So I guess I just need to buckle up and chin up. I'll figure it out. Whatever the others need I'll be there but I just will try to be quieter about it.

But that’s only the bad side of the feelings. And the side that can go straight into the crevices of the purple box.

I will focus on the good things. The real things. The things I WANT to feel and know. Which is that Ca wants to change. That she seems more confident. That she was honest. I hope she was honest. She hugged me so she felt comfortable with me again. That she also wants to be my friend. That she wants to be my friend. That’s all I wanted for the longest. So I can be happy that at least maybe something about what I did worked? I know it mostly came from her but I’m glad I was stubborn and stayed kind so now she feels able to try. I’m excited to see her grow. I just have to remember to be a more normal human. Less about my connection with the Abyss. Let Ir take the lead while Ca and I get to being better. She seems to full heartedly feel much more comfortable with the Abyss when it comes from Ir. I know now to slow it down for her. To not just spout off random facts. To be more human. Let Ir or someone with as much experience like Kine talk or be sourced as a reference.

My excitement makes me look like I’m rubbing it in so I will wait when in group settings. It won’t be forever. And I am just happy to try and make Ca comfortable.

I really am excited she wants to try.

And I’ll cherish that feeling for the rest of the weekend. That hope that things are going to get better. I just had to let it all out here like an outlet. Now I can just be happy. Next up is in fact looking for every Disney movie that’s ever made me cry so I can sob it all out tonight.

[nonbinary-coder]

Success!

I have completed the Dodge the Creeps tutorial in Godot, and thus have made my first fully functional 2D Godot Engine game! :D

there was some sort of issue with my save last night, so that when I reopened the file to keep working on it, it just wouldn't load, so I started over. (I suspect the problem was that I left the program open when I closed my computer, instead of closing it) Didn't take me long to catch back up to where I had been, and this time around I better understood what I was doing.

I made some modifications as I went along, including adjusting the animation switching logic to be smoother, and slowing down the spawn rate of the monsters.

I also figured out how to modify the appearance of the Godot Editor itself! The bounding box in the 2D Scene Editor, which shows you the outline of the game screen, was so faint that I didn't notice it was there at first and then when trying to use it as a reference to place the mob spawn path and to position the HUD bits, I had to squint to see it. Similarly, the comments in the Script Editor screen were hard to read because of their color.

I found the Editor Settings (with the help of someone on reddit who had previously answered a similar question), and changed both those colors. Now they're much easier to see.

Soon, I shall start on coding my own game ideas.

[auntiealiasing]

I think this is called a "bounding box", though I'm not sure if that term also applies to 2D games

Overlays showing the invisible camera movement activation lines in Super Mario Bros. and Super Mario World.

Left: in Super Mario Bros., the camera only moves to the right. As such, if Mario is to the left of the left line, the camera does not move. When he touches the left line, it begins to move but does not match his speed, increasing in speed the closer he is to the right line. Finally, when he touches the right line, the camera matches his speed, preventing him from ever getting further than that point on the screen as long as there is still space for the camera to keep scrolling to the right.

Right: in Super Mario World, the concept is elaborated on in horizontal levels and made double-sided to accommodate both left and right scrolling. Whenever Mario touches a dashed line, the camera moves to align him with the solid line with the same-direction flag (the left solid line when moving right, and vice versa), and keeps up with him as long as he keeps moving.

Main Blog | Twitter | Patreon | Small Findings | Source

[gtsconsultantin]

Video Annotation Services: Transforming Autonomous Vehicle Training

Introduction:

As autonomous vehicles (AVs) progressively Video Annotation Services shape the future of transportation, the underlying technology is heavily dependent on precise and comprehensive datasets. A pivotal element facilitating this advancement is video annotation services. These services enable machine learning models to accurately perceive, interpret, and react to their environment, rendering them essential for the training of autonomous vehicles.

The Importance of Video Annotation in Autonomous Vehicles

Autonomous vehicles utilize sophisticated computer vision systems to analyze real-world data. These systems must be capable of recognizing and responding to a variety of road situations, including the identification of pedestrians, vehicles, traffic signals, road signs, lane markings, and potential hazards. Video annotation services play a crucial role in converting raw video footage into labeled datasets, allowing AI models to effectively "learn" from visual information.

The contributions of video annotation to AV training include:

Object Detection and Classification Video annotation facilitates the identification and labeling of objects such as cars, bicycles, pedestrians, and streetlights. These labels assist the AI model in comprehending various objects and their relevance on the road.

Lane and Boundary Detection By annotating road lanes and boundaries, autonomous vehicles can maintain their designated paths and execute accurate turns, thereby improving safety and navigation.

Tracking Moving Objects Frame-by-frame annotation allows AI models to monitor the movement of objects, enabling them to predict trajectories and avoid collisions.

Semantic Segmentation Annotating each pixel within a frame offers a comprehensive understanding of road environments, including sidewalks, crosswalks, and off-road areas.

Scenario-Based Training Annotated videos that encompass a range of driving scenarios—such as urban traffic, highways, and challenging weather conditions—aid in training AVs to navigate real-world complexities.

The Importance of High-Quality Video Annotation Services

The development of autonomous vehicles necessitates extensive annotated video data. The precision and dependability of these annotations significantly influence the effectiveness of AI models. Here are the reasons why collaborating with a professional video annotation service provider is essential:

Expertise in Complex Situations: Professionals possess a deep understanding of the intricacies involved in labeling complex and dynamic road environments.

Utilization of Advanced Tools and Techniques: High-quality video annotation services employ state-of-the-art tools, such as 2D and 3D annotation, bounding boxes, polygons, and semantic segmentation.

Scalability: As the development of autonomous vehicles expands, service providers are equipped to manage large volumes of data efficiently.

Consistency and Precision: Automated quality checks, along with manual reviews, guarantee that annotations adhere to the highest standards.

How Transforms Video Annotation

At we focus on providing exceptional image and video annotation services specifically designed for the training of autonomous vehicles. Our team merges technical proficiency with advanced tools to generate datasets that foster innovation within the AV sector.

Key Features of Our Offerings:

Tailored annotation solutions to address specific project requirements.

Support for a variety of annotation types, including bounding boxes, 3D point clouds, and polygon annotations.

Stringent quality assurance protocols to ensure data accuracy.

Scalable solutions capable of accommodating projects of any size or complexity.

By selecting you secure a dependable partner dedicated to enhancing the performance of your AI models and expediting the advancement of autonomous vehicles.

The Future of Autonomous Vehicle Training

As the demand for autonomous vehicles Globose Technology Solutions continues to rise, the necessity for accurate and diverse datasets will become increasingly critical. Video annotation services will play a pivotal role in facilitating safer, smarter, and more efficient AV systems. By investing in high-quality annotation services, companies can ensure their AI models are well-prepared to navigate the complexities of real-world environments. The success of your AI initiatives, whether in the realm of self-driving vehicles, drones, or other autonomous systems, heavily relies on video annotation services. Collaborating with specialists such as can help convert unprocessed video data into valuable insights, thereby propelling your innovation efforts.

[svstainless12]

Stainless Steel 202 Sheets Exporters in India

In the realm of stainless steel wonders, Stainless Steel 202 Sheets stand tall as a testament to durability, versatility and flexibility. Crafted with a perfect blend of chromium, nickel and manganese, these cards from Sri Venkatesh Wires and Steels Pvt Ltd are the epitome of quality craftsmanship.

Our stainless steel 202 sheets boast a similar look to the popular A240 stainless steel, yet retain their unique appeal. They exhibit exceptional hardness even in the cold, making them indispensable in limited applications. From kitchen equipment to trains, from cookware to passenger cars, its usefulness knows no bounds.

But like any masterpiece, care is paramount to preserve their integrity. Organic water, acids, alkalis, and salty substances can pose a threat to these leaves. Also, the corrosive fingerprints of polluted air can ruin their shine. Careful storage is key to protecting against such enemies. A regular cleaning ritual combined with a pre-treatment of rust preventatives ensures that these herbs retain their pristine beauty.

At Shree Venktesh Wires and Steels Pvt Limited, we pride ourselves on delivering excellence on every note. Our ASME SA240 Workshop Nr. The 1.4373 pages conform to precise specifications and include detailed mill test reports, giving our customers peace of mind. With our wide range of manufacturing options, customization is effortless and meets needs with ease.

Strength is synonymous with our stainless steel 202 sheets. Their sturdy construction combined with a lightweight design emphasizes durability and durability. Rest assured, our documentation undergoes rigorous testing to maintain the highest standards of quality, ensuring surIn the realm of stainless steel wonders, Stainless Steel 202 Sheets stand tall as a testament to durability, versatility and flexibility. Crafted with a perfect blend of chromium, nickel and manganese, these cards from Sri Venkatesh Wires and Steels Pvt Ltd are the epitome of quality craftsmanship.

Our stainless steel 202 sheets boast a similar look to the popular A240 stainless steel, yet retain their unique appeal. They exhibit exceptional hardness even in the cold, making them indispensable in limited applications. From kitchen equipment to trains, from cookware to passenger cars, its usefulness knows no bounds.

But like any masterpiece, care is paramount to preserve their integrity. Organic water, acids, alkalis, and salty substances can pose a threat to these leaves. Also, the corrosive fingerprints of polluted air can ruin their shine. Careful storage is key to protecting against such enemies. Regular cleaning rituals combined with a pre-treatment of rust preventatives ensure that these herbs retain their pristine beauty.

At Shree Venktesh Wires and Steels Pvt Limited, we pride ourselves on delivering excellence on every note. Our ASME SA240 Workshop Nr. The 1.4373 pages conform to precise specifications and include detailed mill test reports, giving our customers peace of mind. With our wide range of manufacturing options, customization is effortless and meets needs with ease.

Specification/Grades of 202 Stainless Steel Sheets

Thickness  0.3~10.0mm

Width  1000mm, 1219mm, 1500mm, 1800mm, 2000mm, 2500mm, 3000mm, 3500mm, etc

Length  2000mm, 2440mm, 3000mm, 5800mm, 6000mm, etc

Standard  JIS, AISI, ASTM, GB, DIN, EN, etc

Surface  the surface and finish of stainless steel 202 sheet:   2B, 2D, BA, NO.1, NO.4, NO.8, 8K, mirror, checkered, embossed, hair line, sand   blast, Brush, etching, etc

Brand: Indian Origin, European Origin, Japanese Origin, US Origin, Korea Origin, Thailand Origin,   Taiwan Origin

Manufacturer:  POSCO, Aperam, Jindal Stainless, DKC Korea, Thyssenkrup, Baosteel, TISCO, Arcelor Mittal, VDM,   Nippon Metal, Outokumpu

Package  Shrink-wrapped ,Carton boxes,Wooden pallets,Wooden boxes ,Wooden crates Etc.   required.

Export to  India, Indonesia, Malaysia, Singapore, Saudi Arabia, Qatar, UAE, Iran, Iraq, Oman,   South Africa, Nigeria,     Kenya, Italy, UK, Brazil, Russia , Egypt , Kuwait , Taiwan, Australia .

[pixelannotation]

https://pixelannotation.mystrikingly.com/blog/exploring-the-use-of-2d-bounding-boxes-in-retail-and-e-commerce-applications

[hanszoe]

okay i thought "i can use reading lookup" to associate furigana but the entire point of it is that some kanji use different furigana. i could theoretically say the likelihood of this happening twice in a single line is low so we take the only kanji with unrecognized reading and associate them. nope i'll need to do some kind of cursed 2d math again to take the center of bounding box and like... sliide over to the right a little and see what is there. really we should be identifying character by character and handling this process there and doing some kind of intermediary pos identification a la AST lexing (vastly inferior to ANNs i guess?). i have an encroaching sense of doom about unavoidable model training fate

[em2dplatformer]

Adding a HUD to my 2D Platformer

To add a HUD, we need to add a Widget Blueprint. This blueprint allows me to add text and images to the screen. We add this by selecting 'User Interface' when adding in the content drawer and then selecting 'Widget Blueprint'.

A widget allows us to add a HUD by adding 2D elements to the screen of a normally 3D game. This means I can add things like a health bar, score and inventory slots.

For the widget to actually appear on the screen, I need to add it to my character at the 'Event BeginPlay'.

To begin, I added a canvas panel. This is so that the text and images can be placed in the correct corner of the screen.

Next is adding the text.

To edit the text size, you have to use the 'Appearance' section in details, resizing the box does not change the text's size. Colour, letter spacing and font are also found here.

Next, to add an image. Download the desired image as a PNG and drag it into the content drawer. Apply 2D paper settings to the image if the background needs to be removed. Now the image can be dragged from the content drawer into the canvas.

Now the image can be moved, resized and made less opaque.

For everything added to the canvas panel, make sure to anchor them. This prevents them from warping when the size of the window is changed.

Binding:

To get text on the HUD to provide actual information, it needs to be bound to that info. Next to the text part in details, there is an option called 'Bind'. In this option you can either create a new binding or use any ones previously made for this HUD.

When clicked, the bind will first look like this in the graph.

Score:

Now I want this piece of text to tell the player their score, meaning it needs to be bound to the score. This is done by casting to whatever is storing the score value, in my case that's my game instance. From the cast, both 'Get Game Instance' and 'Get Score' are needed.

Next is making it so there's actual text that shows the score. This is done by connecting the 'Game Inst Score' to an 'Append' action followed by a 'To Text (String)' action. This should then connect up to the Return Node along with the 'Exec' link from 'Cast to MyGameInstant'.

This is what the whole code should look like. The bind should be added to a textbox that is separate to the 'Score:' text

The '0' is a separate piece of text to the 'Score:' so that the bind won't affect the rest of the text.

Health:

For the health, I first started with a copy of the score, using 'Get Health' instead of 'Get Score'.

Now, because we don't need to know the exact health and because I prefer it, I'm going to create a health bar to use rather than the standard numbers. For this we first add a 'ProgressBar' into the HUD.

Next is binding the health bar to the players actual health. This is done in a similar way to before, but now requires the players current health to be divided by their full health so that the amount can be calculated by the bar. It also requires the health integer to be converted into a float so that there is an infinite number of answers and won't end up accidentally giving us 0.

[atangledfate]

Maybe he was jumping to conclusions? The thought of being tamed though made him think to much of being in chains again. A creature bound to another and forced to serve under them. It was a soft spot for him considering what he was made for and forced to do in starlines name. He never had the choice to do or be anything else until he broke those chains. No that wasn't accurate was it? Surge broke those chains, and he simply followed along doing as he was told for a long time. either way--- He didn't like the thought of taming Serene, though a mutual friendship did seem to be blossoming. He did care for her, and maybe even was developing feelings for her... it was still an odd thought for him to have or even admit to himself.

His eyes gazed across the many machines with curiosity. He didn't recognize most of them, but that only made his interest in trying them greater. Though hearing Serene rattle off what she enjoyed made him think her tastes were closer to Surge. Funny to think that despite being so fast she still loved racing games to death. She really enjoyed those gear races as if some part of her was drawn to it. Maybe it was related to those memories resurfacing.

" I like a good shooter or racer as well... kart racers are probably my favorite, just for the simplicity of it. Though i tend to stay away from shooters... seems to trigger to much of my battle program... last time i ripped the gun off the arcade box... "

he fidgeted at the memory of losing his cool and his program taking over. It wasn't pretty, and it took surge to calm him down afterwards but, that was a long time ago. He was much more controlled now then he was before they left mobius.

He was startled by her pulling him along as he was just thinking about the incident. But being brought over to the arcade cabinet instantly had his interest. He seemed to watch the game play in the background a moment just getting a feel for the type of fighter it was. He typically played 3D fighters but he did love a good 2D fighter on occasion. This reminded him a bit of Allyway Fighter 2 back on mobius-- course he found the adaptation of quadruped characters pretty unique.

" Interesting... they are a colorful cast of characters. Reminds me a bit of Sonic and his friends... funny that tails never made something like this for fun... but let's see how i fair... "

He paused looking around for a change machine or, coins. Honestly he was so use to surge just stealing the coins from the machines with her electro-magnokinesis.

" mmm coins? where do i... get those "

She grimaced softly as she wandered into the arcade with him, "Yeah... Perhaps tame is a bit too strong of a word..."

Even so, it takes a great deal of trust to allow someone to be that close to 'monster.' And him? Kistunami is definitly on the list of potential riders. In any case, the sights and sounds of the arcade drew her attention. Did they get a new game?

She pratically bounced in place as she answers his question, "I prefer shooters amd racing games. The former because of the precision it takes to get a good score, and the latter because... Well, its fun to go fast and outsmart other racers with clever uses of boost and items."

That, and the fact that games inspired by other races and thier histories have been cropping up al ot lately have been breathing new life into the Arcade.

There, they would see ponies, dragons, gryphons, and a number of other folk milling about, enjoying thier time. There was even a changedling playing Prance Prance Evolution in the corner.

Serene would grab Kit by the hand and lead him over to the opposite corner of the arcade. While it wasn't exactly her thing, she was curious to see just how good he was at playing his favorite games. Fighting is Magic doesn't have many cabinets, but the Princess made sure to preserve one right here in her own home town.

"Fun fact about this game, the characters are all based on the Princess and her friends when they were all younger... This game is nearly two hundred years old if you can beleive it. Beyond vintage."

[fmpyear2kb]

Hastily developing the fuelling level

It was obvious that the fuelling level wasn't the main focus of Da'Car 2, though it will be developed further in the future. To begin this level I inserted two cubes, one being the desert landscape, and the other being the concrete the gas station would be on.

From here I quickly dragged Da'Car in and a few zombie spawners from the first game.

It was important to understand how the layout of the level would affect the gameplay, so I took a slight break and did a little more development on the car chase level. When I came back to this one, I added the bounds box after considering a few options.

These boxes essentially creates the play area where both the player and zombies are able to move around. A navmesh was also added so that the AI actually could navigate the area, depending on the size of the mesh.

The green area is the navmesh, allowing the AI to move around the box as I see fit. With the basics complete, it was time to insert the gas station model I found online.

However, the textures imported with the model were surprisingly too detailed for the texture sample in unreal to withstand. The model was perfect, though despite the materials being unusable, the model stuck out like a sore thumb as it was a regular 3d object in a mainly 2d based art styled game. Here's the station in the editor with no textures.

After the voxel gas station was modelled, it was swiftly inserted into the game.

Due to how the camera worked, I couldn't include a roof to the station, instead there was a floating cube that was hidden in game to cast a shadow as if you were underneath the station.

This is where it was left until we received feedback from people playtesting saying that there were too many zombies and the car was becoming an obstacle. Firstly I deleted some spawners and decreased the maximum number of randomised zombies that could spawn per wave.

I reduced the spawners from 7, to 3 and decreased the maximum range from 9, to 6. These changes helped balance the amount of zombies spawned, allowing the player to not be overrun by enemies too easily. The collision of the car was also changed so you could shoot through it as the car being an obstacle made it difficult to kill zombies on the opposite side.

With the collision now ignoring the world dynamic collisions channel, the bullet could now pass through the actor but still block the player from walking through.

These changes would conclude the development on this level, there's a lot more that needs to be done to it, however I need to make the hard decision on what needs to be cut and what needs to be added due to the small amount of time left.

[gtsconsultantin]

Struggling with Data Labeling? Try These Image Annotation Services

Introduction:

In the era of artificial intelligence and machine learning,Image Annotation Services data is the driving force. However, raw data alone isn’t enough; it needs to be structured and labeled to be useful. For businesses and developers working on AI models, especially those involving computer vision, accurate image annotation is crucial. But data labeling is no small task. It’s time-consuming, resource-intensive, and requires a meticulous approach.

If you’ve been struggling with data labeling, you’re not alone. The good news is that professional image annotation services can make this process seamless and efficient. Here’s a closer look at why data labeling is challenging, the importance of image annotation, and the best services to help you get it done.

The Challenges of Data Labeling

Time-Consuming Process

Labeling thousands or even millions of images can take an enormous amount of time, delaying project timelines and slowing innovation.

High Cost of In-House Teams

Building and maintaining an in-house team for data labeling can be costly, especially for small and medium-sized businesses.

Need for Precision

AI models require accurate and consistent labels. Even minor errors in annotation can significantly impact the performance of your AI systems.

Scaling Issues As your dataset grows, so do the challenges of managing, labeling, and ensuring quality control at scale.

The Importance of Image Annotation

Image annotation involves adding metadata or labels to images, helping AI systems understand what’s in a picture. These annotations are used to train models for tasks such as:

Object detection

Image segmentation

Facial recognition

Autonomous driving systems

Medical imaging analysis

Without proper annotation, AI models cannot interpret visual data effectively, leading to inaccurate predictions and unreliable outputs.

Top Image Annotation Services to Streamline Your Projects

If you’re ready to take your AI projects to the next level, here are some top-notch image annotation services to consider:

Offers a range of high-quality image and video annotation services tailored to various industries, including healthcare, retail, and automotive. With a focus on precision and scalability, they ensure your data labeling needs are met efficiently.

Key Features:

Bounding boxes, polygons, and semantic segmentation

Annotation for 2D and 3D data

Scalable solutions for large datasets

Affordable pricing plans

Scale AI

Scale AI provides a comprehensive suite of data annotation services, including image, video, and text labeling. Their platform combines human expertise with machine learning tools to deliver high-quality annotations.

Key Features:

Rapid turnaround times

Detailed quality assurance

Customizable annotation workflows

Labelbox

Labelbox is a popular platform for managing and annotating datasets. Its intuitive interface and robust toolset make it a favorite for teams working on complex computer vision projects.

Key Features:

Integration with ML pipelines

Flexible annotation tools

Collaboration-friendly platform

CloudFactory

CloudFactory specializes in combining human intelligence with automation to deliver precise image annotations. Their managed workforce is trained to handle intricate labeling tasks with accuracy.

Key Features:

Workforce scalability

Specialized training for annotators

Multilingual support

Amazon SageMaker Ground Truth

Amazon’s SageMaker Ground Truth is a powerful tool for building labeled datasets. It uses machine learning to automate annotation and reduce manual effort.

Key Features:

Active learning integration

Pay-as-you-go pricing

Automated labeling workflows

Why Choose Professional Image Annotation Services?

Outsourcing your image annotation tasks offers several benefits:

Expertise: Professionals have the tools and experience to deliver precise annotations.

Efficiency: Save time and focus on your core business activities while experts handle the data labeling.

Scalability: Easily scale your annotation efforts as your dataset grows.

Cost-Effectiveness: Eliminate the need for in-house teams and costly software investments.

Conclusion

Data labeling doesn’t have to be a bottleneck for your AI projects. By leveraging professional image annotation services like Globose Technology Solutions and others, you can ensure your models are trained on high-quality, accurately labeled datasets. This not only saves time and resources but also enhances the performance of your AI systems.

So, why struggle with data labeling when you can rely on experts to do it for you? Explore the services mentioned above and take the first step toward seamless, efficient, and accurate image annotation today.