Case study
RU-SAFE Campus Security System
A CV-based security prototype that almost won a campus-wide competition.
- Timeline
- Oct 2025 — Mar 2026
- Stack
- Product Strategy, Python (OpenCV, YOLO, MediaPipe), Computer Vision, Technical Pitching
- 97.7%Decrease in notification time
- 80%Of student body that would benefit
- 10+11pm phone calls working on our pitch, code, or other things
“If a crime were to happen to you right now, nobody would know until a day later.” - our opening pitch, and the reason why we designed a product to cut response time.
Among many iterations of our pitch, the opening line I remember most was “Do you feel safe right now?”. For obvious reasons, we scrapped it, but it’s a fond reminder of the many many hours I spent writing our pitches, then saying them back to myself with different deliveries, trying to make every single word matter as much as possible.
The Idea
I will be honest, I was not passionate about this project at first. I got looped into doing the Verizon Smart Campus Competition by a few friends, but I wasn’t all in yet. Not nearly. I came up with a few ideas. I listened to a few of my teammates’ ideas. All overdone. Bus optimization algorithms, point-cloud real estate mapping, all very useful ideas, but their scope was so limited. It was all Rutgers-based, which wasn’t a problem, but it was Rutgers-only, which was. Then, a friend was inspired by a Rutgers crime alert (an email notification, often delayed, to let students know about recent crimes on campus), and suggested that we come up with a way to fix that.
I’ll be the first to admit that it took me quite a while to really get on board with it. I thought, “Security? But that’s not cool. It doesn’t have a nice ring to it that sounds impressive.” But that was the idea we decided on, and we went into it. We came up with a name, an idea of our scope, and an outline for how we were going to pitch it. We recorded a 2 minute video, then submitted it in about November, then I thought that was that and I’d never hear about it again.
Round 2
A few months of nothingness later, we were notified that we were moving on to the second round of the competition. It seemed like our idea had finally gotten some wheels. At this point, I realized that if we were to win this competition, I had to step up and become the group lead, and make sure everyone was doing what they did best. Our efforts at this point were haphazard but spirited, and it took some organizing to get everyone to help out where they could help best. Eventually, we made it to the second round event, which was a poster-based pitch. Our two-minute pitches were rehearsed down to a T, we wore matching red and black outfits, and we had a somewhat functioning mockup of a front-end. We made it through, and so did some of the teams we made friends with, so we were happy.
The Glorious Pitch
We were then told we had four weeks until the final pitch. A ten minute presentation with five minutes at the end for questions. The judges would be Verizon professionals, Rutgers deans, and otherwise very smart, very experienced people. The stakes were on. We made the executive decision to step up our matching red and black outfits to include suit jackets. Some of us focused on making a functioning front-end. I made a functioning back-end. This back-end could use your device’s camera to watch events unfold on the screen, then analyze it to detect weapons and aggressive motions. That alone took me two weeks. After that, we started working on our demo. We knew that if we wanted to win, it would have to be electric.
And it was. We knew our pitch was going to be the winning pitch. We were casual and professional, funny but serious, and after several practice runs, we were good. The day of, we did a few final tweaks to our demo, making sure that the parameters for our prototype were tuned so when my teammate was going to stab me with a fake knife mid-presentation, our algorithm’s alarm bells would start ringing and everyone in the room would get a notification less than thirty seconds later.
We asked the hosts if we could go last, because our pitch required a somewhat complicated setup with my phone and laptop on different sides of the stage. That, and we wanted our presentation to be the one that stuck with everyone. Our introduction, minus a small hiccup where a teammate blanked for a split second, was amazing. People were laughing along with us, and when my teammate took out the knife to stab me (again, it’s a fake knife, don’t worry), the anticipation in the room was exactly what we were going for: electric.
Then it failed. The camera couldn’t start, no matter how much we tried. Maybe because it was too far, or that I received a phone call mid-demo. Whatever it was, our demo completely collapsed, and we were forced to resort to the backup pictures on our slide deck.
The Many Takeaways
We recovered as best we could, but with a presentation that’s intrinsically reliant on your demo, there’s only so much you can do without said demo. And so, when the winners were announced, we were disappointed, if not terribly surprised, that we weren’t among them.
Despite not winning, though, this was one of my most impactful projects ever. I learned a great deal about developing a product, both from the technical side and from the “is this actually something people want” side. If I had to do it differently, I think I would’ve done some more research in the beginning and really understood what students actually wanted for campus security, and if Big Brother Lite was a part of that. Despite that, I learned a lot, and it’s probably one of the biggest projects that secured me my internship and first professional experience.
Among many iterations of our pitch, the opening line I remember most was “Do you feel safe right now?”. For obvious reasons, we scrapped it, but it’s a fond reminder of the many many hours I spent writing our pitches, then saying them back to myself with different deliveries, trying to make every single word matter as much as possible.
The Idea
I will be honest, I was not passionate about this project at first. I got looped into doing the Verizon Smart Campus Competition by a few friends, but I wasn’t all in yet. Not nearly. I came up with a few ideas. I listened to a few of my teammates’ ideas. All overdone. Bus optimization algorithms, point-cloud real estate mapping, all very useful ideas, but their scope was so limited. It was all Rutgers-based, which wasn’t a problem, but it was Rutgers-only, which was. Then, a friend was inspired by a Rutgers crime alert (an email notification, often delayed, to let students know about recent crimes on campus), and suggested that we come up with a way to fix that.
I’ll be the first to admit that it took me quite a while to really get on board with it. I thought, “Security? But that’s not cool. It doesn’t have a nice ring to it that sounds impressive.” But that was the idea we decided on, and we went into it. We came up with a name, an idea of our scope, and an outline for how we were going to pitch it. We recorded a 2 minute video, then submitted it in about November, then I thought that was that and I’d never hear about it again.
Round 2
A few months of nothingness later, we were notified that we were moving on to the second round of the competition. It seemed like our idea had finally gotten some wheels. At this point, I realized that if we were to win this competition, I had to step up and become the group lead, and make sure everyone was doing what they did best. Our efforts at this point were haphazard but spirited, and it took some organizing to get everyone to help out where they could help best. Eventually, we made it to the second round event, which was a poster-based pitch. Our two-minute pitches were rehearsed down to a T, we wore matching red and black outfits, and we had a somewhat functioning mockup of a front-end. We made it through, and so did some of the teams we made friends with, so we were happy.
The Glorious Pitch
We were then told we had four weeks until the final pitch. A ten minute presentation with five minutes at the end for questions. The judges would be Verizon professionals, Rutgers deans, and otherwise very smart, very experienced people. The stakes were on. We made the executive decision to step up our matching red and black outfits to include suit jackets. Some of us focused on making a functioning front-end. I made a functioning back-end. This back-end could use your device’s camera to watch events unfold on the screen, then analyze it to detect weapons and aggressive motions. That alone took me two weeks. After that, we started working on our demo. We knew that if we wanted to win, it would have to be electric.
And it was. We knew our pitch was going to be the winning pitch. We were casual and professional, funny but serious, and after several practice runs, we were good. The day of, we did a few final tweaks to our demo, making sure that the parameters for our prototype were tuned so when my teammate was going to stab me with a fake knife mid-presentation, our algorithm’s alarm bells would start ringing and everyone in the room would get a notification less than thirty seconds later.
We asked the hosts if we could go last, because our pitch required a somewhat complicated setup with my phone and laptop on different sides of the stage. That, and we wanted our presentation to be the one that stuck with everyone. Our introduction, minus a small hiccup where a teammate blanked for a split second, was amazing. People were laughing along with us, and when my teammate took out the knife to stab me (again, it’s a fake knife, don’t worry), the anticipation in the room was exactly what we were going for: electric.
Then it failed. The camera couldn’t start, no matter how much we tried. Maybe because it was too far, or that I received a phone call mid-demo. Whatever it was, our demo completely collapsed, and we were forced to resort to the backup pictures on our slide deck.
The Many Takeaways
We recovered as best we could, but with a presentation that’s intrinsically reliant on your demo, there’s only so much you can do without said demo. And so, when the winners were announced, we were disappointed, if not terribly surprised, that we weren’t among them.
Despite not winning, though, this was one of my most impactful projects ever. I learned a great deal about developing a product, both from the technical side and from the “is this actually something people want” side. If I had to do it differently, I think I would’ve done some more research in the beginning and really understood what students actually wanted for campus security, and if Big Brother Lite was a part of that. Despite that, I learned a lot, and it’s probably one of the biggest projects that secured me my internship and first professional experience.
Among several iterations of our first live pitch, the opening line I remember most was “Do you feel safe right now?”. For obvious reasons, we scrapped it, but it’s a fond reminder of the many, many hours I spent writing our pitches, then saying them aloud to myself or my teammates with different deliveries, trying to make every single word matter as much as possible.
The Verizon Smart Campus Competition was a campus-wide competition where our task was to solve some sort of problem on campus with a solution that was scalable. There were three rounds. The first was a two-minute video, the next was a three-minute poster showcase pitch, and the final round was a ten-minute presentation with a five-minute Q&A. Our final round presentation was something I will never forget.
The Idea
I will be honest, I was not passionate about this project at first. When I first got looped into doing it, I wasn’t all in just yet. Not nearly. I came up with a few ideas. I listened to a few of my teammates’ ideas. All overdone. Bus optimization algorithms, point-cloud real estate mapping, all very useful, but such limited scope. It was all Rutgers-based, which wasn’t a problem, but it was Rutgers-only, which was. I knew from previous experience that if you want to win, you need to have something that is not only useful, but cool. Then, a friend was inspired by a Rutgers crime alert (an email notification, often delayed, to let students know about recent crimes on campus), and suggested that we come up with a way to fix that.
I’ll be the first to admit that it took me quite a while to really get on board with it. I thought, “Security? But that’s not cool. It doesn’t have a nice ring to it that sounds impressive.” But that was the idea we voted on, and we went into it. We came up with a name, an idea of our scope, and an outline for how we were going to pitch it.
The day of our Round 1 video recording, it was a chilly November day. Naturally, our day involved several takes of filming a fake robbery, just to drive our point home. We talked about the problem on campus, how crime alerts are often so delayed that they’re practically useless. We talked about why it was happening, and what our (AKA the best) way to fix it was: AI-powered cameras that can scan for crimes and instantly take action when one happens. We talked about how this wasn’t just a solution for Rutgers, but a scalable system that could be deployed at college campuses, cities, and just about anywhere where seconds matter. We submitted our two-minute video, then I thought that was that and I’d never hear about it again.
Round 2
A few months of nothingness later, we were notified that we were moving on to the second round of the competition. It seemed like our idea had finally gotten some wheels. At this point, I realized that if we were to win this competition, I had to step up and become the group lead, and make sure everyone was doing what they did best. Our efforts at this point were haphazard but spirited, and it took some organizing to get everyone to help out where they could help best. Once people knew how they could help, we started our poster. We collected some data from a student base (my friends) so we could narrow down our scope and work on the most frustrating problems for students. What we found from research and student opinions was that it’s the information gap that causes the biggest problem. If the authorities only find out a day after a crime happens, what good does it do if they can get there in 20 seconds or 20 minutes? So, we focused on reducing the crime notification time.
At this point, we didn’t know exactly how we were going to build it, but after some research, I found that MediaPipe and YOLO were two open-source models that would help us detect crimes. How exactly we’d do that, we didn’t know, but we knew it was possible, so we put it on our poster. We also made a user interface. Thanks to the genius of my teammate and the utility of Railway, in about a week, he was able to deploy a mostly working front-end that showed a mockup of how we wanted our interface to look.
Eventually, we made it to the second round event, which was a poster-based pitch. Our two-minute pitches were rehearsed down to a T (then we realized we actually had three minutes and didn’t have to talk faster than Eminem), we wore matching red and black outfits, and we were ready to field questions. Some of the questions left us reeling, but from each of those questions, we learned something and planned to implement it later. We made it through, and so did most of the teams we made friends with, so we were happy.
The Build(up)
We were then told we had four weeks until the final pitch. A ten-minute presentation with five minutes at the end for questions. The judges would be Verizon professionals, Rutgers deans, and otherwise very smart, very experienced people. The stakes were on. We made the executive decision to step up our matching red and black outfits to include suit jackets. Some of us focused on making a functioning front-end. I made a functioning back-end.
The idea for the back-end was to demonstrate a few things: first, the type of code that would be deployed onto edge compute devices to capture data from cameras. We wired my phone’s camera to my computer and used the open-source software we’d decided on to watch events unfold on the screen, then analyze them to detect weapons and aggressive motions. Building that out alone took me two weeks. Refining it, forget about it. The second thing we wanted to demonstrate was that AI models can be wrong. If the model detects a false positive, we wanted to include a human review layer. Despite slowing us down from milliseconds to minutes, we decided the trade-off was worth it to maximize accuracy. So to that end, we set up a popup that would simulate what a security personnel would see on their screens. This human reviewer dashboard would show the crimes detected and where they were, and based on the real footage, the reviewer would accept it (which would send a notification) or reject it.
After we wired the front-end and the back-end together using a makeshift Google Sheets database, we started working on our pitch, which revolved around our demo. We knew that if we wanted to win, it would have to be electric.
The Glorious Pitch
And it was. We knew our pitch was going to be the winning pitch. We were casual and professional, funny but serious, and after several practice runs, we were good. The day of, we did a few final tweaks to our demo, making sure that the parameters for our prototype were tuned so when my teammate was going to stab me with a fake knife mid-presentation, our algorithm’s alarm bells would start ringing and everyone in the room would get a notification less than thirty seconds later.
We asked the hosts if we could go last, because our pitch required a somewhat complicated setup with my phone and laptop on different sides of the stage. That, and we wanted our presentation to be the one that stuck with everyone. Our introduction, minus a small hiccup where a teammate blanked for a split second, was amazing. People were laughing along with us, and when my teammate took out the knife to stab me (again, it’s a fake knife, don’t worry), the anticipation in the room was exactly what we were going for: electric.
Then it failed. The camera couldn’t start, no matter how much we tried. Maybe because it was too far and my computer couldn’t pick up the signal, or that I received a phone call mid-demo. Whatever it was, our demo completely collapsed, and we were forced to resort to the backup pictures on our slide deck.
The Many Takeaways
We recovered as best we could, but with a presentation that’s so reliant on your demo, there’s only so much you can do without said demo. And so, when the winners were announced, we were disappointed, if not terribly surprised, that we weren’t among them.
Despite not winning, though, this was one of my most impactful projects ever. I learned a great deal about developing a product, both from the technical side and from the “is this actually something people want” side. Despite not feeling too strongly about the idea at first, it grew on me to the point where I was putting in hours of work, occasionally missing a few classes, and otherwise devoting huge amounts of time to this. The late phone calls, the post-defeat Taco Bell session, and our multiple pitches are some of the most grueling, yet fun working sessions I’ve had.
If I had to do it differently, I think I would’ve done some more research in the beginning and really understood what students actually wanted for campus security, and if Big Brother Lite was truly a part of that. From a technical standpoint, I think our model could’ve used a lot more work too. In spite of all that, I learned a lot about how to take an idea from nothing to something, and it’s probably one of the biggest projects that secured me my first professional experience.
Among several iterations of our first live pitch, the opening line I remember most was “Do you feel safe right now?”. For obvious reasons, we scrapped that, but it’s a fond reminder of the many, many hours I spent writing our pitches, then saying them aloud to myself or my teammates with different deliveries, trying to make every single word matter as much as possible.
The Verizon Smart Campus Competition was a campus-wide competition where our task was to solve some sort of problem on campus with a solution that was scalable. There were three rounds. The first was a two-minute video, the next was a three-minute poster showcase pitch, and the final round was a ten-minute presentation with a five-minute Q&A. Our final round presentation was something I will never forget.
The Idea
I will be honest, I was not passionate about this project at first. When I first got looped into doing it, I wasn’t all in just yet. Not nearly. I came up with a few ideas. I listened to a few of my teammates’ ideas. All overdone. Bus optimization algorithms, point-cloud real estate mapping, all very useful, but such limited scope. It was all Rutgers-based, which wasn’t a problem, but it was Rutgers-only, which was. I knew from previous experience that if you want to win, you need to have something that is not only useful, but cool. Then, a friend was inspired by a Rutgers crime alert (an email notification, often delayed, to let students know about recent crimes on campus), and suggested that we come up with a way to fix that.
I’ll be the first to admit that it took me quite a while to really get on board with it. I thought, “Security? But that’s not cool. It doesn’t have a nice ring to it that sounds impressive.” But that was the idea we voted on, and we went into it. We came up with a name, an idea of our scope, and an outline for how we were going to pitch it.
The day of our Round 1 video recording, it was a chilly November day. Naturally, our day involved several takes of filming a fake robbery, just to drive our point home. We talked about the problem on campus, how crime alerts are often so delayed that they’re practically useless. We talked about why it was happening, and what our (AKA the best) way to fix it was: AI-powered cameras that can scan for crimes and instantly take action when one happens. We talked about how this wasn’t just a solution for Rutgers, but a scalable system that could be deployed at college campuses, cities, and just about anywhere where seconds matter. We submitted our two-minute video, then I thought that was that and I’d never hear about it again.
Round 2
A few months of nothingness later, we were notified that we were moving on to the second round of the competition. It seemed like our idea had finally gotten some wheels. At this point, I realized that if we were to win this competition, I had to step up and become the group lead, and make sure everyone was doing what they did best. Our efforts at this point were haphazard but spirited, and it took some organizing to get everyone to help out where they could help best. Once people knew how they could help, we started our poster. We collected some data from a student base (my friends) so we could narrow down our scope and work on the most frustrating problems for students. From this, we found that the biggest problems people had with campus security were that crime notification time was too long. When we investigated further and scraped our previous emails to find every crime alert from the past year, we found that the average time to notify students from when the crime occurred was 21.4 hours. And so we realized that the information gap was the true culprit. If the authorities only find out a day after a crime happens, what good does it do if they can get there in 20 seconds or 20 minutes? So, we focused on reducing the crime notification time.
Our plan was to use the existing security infrastructure on Rutgers’ campus and the City of New Brunswick, wiring the cameras to IoT gateways to retrofit them with the tech we needed. That way, we could keep our implementation low-cost but maximize our reach.
At this point, we didn’t know exactly how we were going to build it, but after some research, I found that MediaPipe and YOLO were two open-source models that would help us detect crimes. We also knew that we wanted to detect objects, specifically weapons, as an added layer. This meant we’d have to train a model to recognize weapons based on datasets. How exactly we’d do that, we didn’t know, but we knew it was possible, so we put it on our poster. We also made a user interface. Thanks to the genius of my teammate and the utility of Railway, in about a week, he was able to deploy a mostly working front-end that showed a mockup of how we wanted our interface to look.
Eventually, we made it to the second round event, which was a poster-based pitch. Our two-minute pitches were rehearsed down to a T (then we realized we actually had three minutes and didn’t have to talk faster than Eminem), we wore matching red and black outfits, and we were ready to field questions. Some of the questions left us reeling, but from each of those questions, we learned something and planned to implement it later. We made it through, and so did most of the teams we made friends with, so we were happy.
The Build(up)
We were then told we had four weeks until the final pitch. A ten-minute presentation with five minutes at the end for questions. The judges would be Verizon professionals, Rutgers deans, and otherwise very smart, very experienced people. The stakes were on. We made the executive decision to step up our matching red and black outfits to include suit jackets. Some of us focused on making a functioning front-end. I made a functioning back-end. Our pipeline is below, for some context.
- Camera
- IoT gateway
- Detection layerYOLOMediaPipe
- Incident flagging
- Human reviewConfirmReject
- AlertSMS / emailDashboard
The back-end was quite interesting. It was designed to receive the data from the cameras (though I just used my phone for the sake of the demo) and use AI to analyze the video. On one side, I developed my own way of recognizing a crime based on pose intent. MediaPipe is a pose estimation software, which means it can capture a video and understand where someone’s limbs are, and can often even go down to the fingers and facial features if the camera is high-resolution enough. I decided to use the body’s overall pose to determine criminal intent based on a heuristic scoring system. Based on a person’s quickness in moving, their leg positioning, their arms moving, and a few other criteria, I could tell that they were lunging towards someone else.
The other side of the back-end involved detecting objects. This one was harder, because crimes often happen at night, when lighting is poor. So, I looked online for free datasets I could use to train a YOLOv8 model. I started with a knife, since that seemed to be relatively simple and it tied into an idea I had for the demo. The data I found wasn’t enough, so I had to supplement it with several images of kitchen knives from my own house as well as the knife we were going to use in our demo, until we were at about 10,000 training images. Due to limited time and being limited to my computer’s power, I was able to eventually get a model that could detect the knife at certain angles. To get that to work, I had to play around with confidence thresholds until the threshold was at a measly 29% to detect the knife. For the demo, it would do, but I knew that for a real-world scenario, I’d need a much more powerful model to get it done.
The next part of our data pipeline involved a human reviewer. We knew AI models could be wrong. Even if we had unlimited power at our disposal, we didn’t want to risk a model automatically detecting a false positive and notifying a student body of 40k+ people of a false alarm. This meant we had to include a human review layer. Despite slowing us down from milliseconds to minutes, we decided the trade-off was worth it to maximize accuracy. So we set up a popup that would simulate what a Rutgers security officer would see on their screens. Once our software flagged a crime, it would pop up on our human reviewer dashboard. After some feedback in Round 2, we even implemented a way to subconsciously reduce bias by showing the MediaPipe stick figures first, then allowing the reviewer to confirm it as a crime or move to a second stage where they could see the actual footage and again confirm it as a crime or deny it.
Once the human reviewer confirmed the crime, we wanted to send out Rutgers crime notifications to everyone. In the real world, we knew we’d have to do that through Rutgers, but we couldn’t resist the idea of having our demo and actually having everyone’s phones go off with a crime alert. So, I used my Google Sheets API key to pull data from a spreadsheet linked to a Google form that we’d give out for everyone to put their information in. Once it pulled the data, our prototype would use Twilio and SMTP to send out a crime alert populated with a short description of the crime and other data pulled from the footage. All in all, we were able to cut alert time from 21.4 hours with the legacy model to under 30 minutes with our product.
We used a makeshift Google Sheets database to push crime alert information, from which my teammate pulled information to populate a map dashboard. This user dashboard was a simple web app that was meant to show recent crime alerts, their severity score, and an expanding bubble to show the affected areas over time (to account for how someone could be on the run). Overall, it was a neat, simple UI that showed users what they needed to see without it being excessive. Our eventual plan was to integrate it with Rutgers’ app for students for simplicity.
Once we wired together our entire data pipeline, we started working on our pitch, which revolved around our demo. We knew that if we wanted to win, it would have to be electric.
The Glorious Pitch
And it was. We knew our pitch was going to be the winning pitch. We were casual and professional, funny but serious, and after several practice runs, we were good. The day of, we did a few final tweaks to our demo, making sure that the parameters for the model were tuned so it would detect the fake knife in my friend’s hand that he was planning on stabbing me with, without detecting everything else in the room. Then, the algorithm’s alarm bells would start ringing, and everyone’s phones would go off at once.
We asked the hosts if we could go last, because our pitch required a somewhat complicated setup with my phone and laptop on different sides of the stage. That, and we wanted our presentation to be the one that stuck with everyone. Our introduction, minus a small hiccup where a teammate blanked for a split second, was amazing. People were laughing along with us, and when my teammate took out the knife to stab me (again, it’s a fake knife, don’t worry), the anticipation in the room was exactly what we were going for: electric.
Then it failed. The camera couldn’t start, no matter how much we tried. Maybe because it was too far and my computer couldn’t pick up the signal, or that I received a phone call mid-demo. Whatever it was, it wasn’t something we had extensively tested, and so our demo completely collapsed, and we were forced to resort to the backup pictures on our slide deck.
The Many Takeaways
We recovered as best we could, but with a presentation that’s so reliant on your demo, there’s only so much you can do without said demo. And so, when the winners were announced, we were disappointed, if not terribly surprised, that we weren’t among them.
Despite not winning, though, this was one of my most impactful projects ever. I learned a great deal about developing a product, both from the technical side and from the “is this actually something people want” side. I trained an AI model, which sounds absolutely insane to say. I made my own way of detecting intent based on body language (while standing on the shoulders of giants). And despite not feeling too strongly about the idea at first, it grew on me to the point where I was putting in hours of work, occasionally missing a few classes, and otherwise devoting huge amounts of time to this. The late phone calls, the post-defeat Taco Bell session, and our multiple pitches are some of the most grueling, yet fun working sessions I’ve had.
If I had to do it differently, I think I would’ve done some more research in the beginning and really understood what students actually wanted for campus security, and if Big Brother Lite was truly a part of that. From a technical standpoint, I think our model could’ve used a lot more work too in terms of training, and I know we should have tested things much more extensively than we did to account for every edge case.
In spite of all that, I learned a lot about how to take an idea from nothing to something, and it’s probably one of the biggest projects that secured me my first professional experience.