Nexar

Turn your iPhone/Android into a community-based AI dashcam

Featured

Discussion

most upvoted
H
Chris MessinaΒ β€” πŸ† PH Community Member of the Year!
These guys were featured 10 months and it looks like they've made considerable progress since then β€” launching their iOS and Android (and getting featured by Apple in the App Store!).

Curious what the team can tell us about their progress and what they've learned since then.
M
eran shirΒ β€” Co-founder & CEO, Nexar
Hey @chrismessina, thanks for hunting us. Happy to share insights from our evolution thus far. We started Nexar to build a realtime, over-the-top vehicle network that can detect dangerous situations and prevent accidents. Our basic assumption is that all we need to do to prevent most of the car collisions in the world is to provide drivers with more reaction time. To do that, we need to understand what every vehicle sees/'feels' and connect drivers and vehicles in a realtime network that senses danger and warns drivers well beyond their line of sight.

So for us, the most exciting milestone in the last year happened less than two weeks ago, on Nov 15th, when we launched our Forward Collision Warning (FWD) and opened our realtime vehicle network (read more here: http://bit.ly/2gvZciU). Nexar users now using iPhone 6S/SE/7 (weaker models will come later) are getting warned in realtime when they get into a situation that may turn into a collision or a pile-up. Just to provide perspective, the concept of V2V existed in labs for over 1 5 years, but the Nexar V2V network is the first one deployed commercially at a city-scale level. If you drive with Nexar today in Manhattan or SF you should be constantly connected to dozens of other vehicles sharing realtime warnings. It's never been done before at this scale, and we still have a lot to learn as a result, but it's damn exciting. What's more, (sorry for the technical detour) we have shown that we can do over-the-top V2V at latencies and rates that were considered Wifi/DSRC-only realm (below 100ms). We are more convinced than ever that our approach is the right one going forward.

You asked about things that we learned, so here are a few:
1. Starting with a focus on pro drivers was a really great move. Uber/Lyft drivers, fleet drivers, truckers. They have a hard, complex life, and they need all the help they can get. Providing them with features like accident reconstruction went such a long way and actually made a difference in their lives. Had we started with the general population these features would've been diluted. Natural niches are great.

2. The world is big and strange - with Millions of driving hours collected, and almost a Million extreme events analyzed, (think accidents, red-light runnings, near-misses with cows in India, and anything in between), what's clear is that the level of richness that exists on the road is vast and unique. We are very bullish on autonomous vehicles, but creating a full, level 5, national-level autonomous vehicle is not going to be trivial.

3. The smartphone is an amazing platform, but you need to be willing to get physical. We needed to get to really high accuracy of location and trajectory, both of our vehicle and the vehicles it 'sees' nearby. Typical telematics solutions just don't cut it. The smartphone have amazing sensors embedded in them today, but some of the manufactures, especially on the Android side, are playing fast and loose when it comes to sensors data quality, so if you really want to gauge the world, you need to be willing to invest a lot on boring subjects such as sensor calibration.

4. Deep learning is our age's Archimedean point. It is amazing the kind of challenges we manage to tackle in short amount of time. But the lever is vast data sets. Without it, you've got nothing. Today, I can reliably say that we, with our small team, can create better quality models and features than many established incumbents just because we got way larger and richer data sets.

5. Doing vision and AI on a smartphone, while possible, is still damn tough. Heating is an issue we had to deal with, dropped frames, and quite a few other issues. When you have an Uber driver using Nexar for 10 hours a day in Sunnyvale or Manhattan in August, you really need to be mindful of such issues.

Ok, this is getting too long, so I'll stop here, but happy to answer more questions below. I'll just end by saying we are working hard on realizing this vision of 'air traffic control for the road' and getting rid of car collisions. In fact, we think it is offensive that in 2016 there are still so many car collisions and the number actually climbs year over year. We are determined to change that.
Honey Raj VarmaΒ β€” Raj Ventures, R&Ds, Internetechs
@eranshir @chrismessina Eran, I love what you're doing and the cause it supports - truly commendable. Perhaps an oversight of sorts, but many users who keep their smartphone cameras on for a while experience not only battery drainage but phone heating. While the charging is not an issue, multi-processes running + camera being on sounds like the device must get hot rather quickly. How do you plan to tackle that challenge?
Jason ShultzΒ β€” Senior Software Engineer, 3form
I've tried this app on both android (Note 7 and Pixel) and iOS (iphone 6s). It works beautifully by itself on iOS except you can't use it with maps unless it's in the foreground. On Android it works great both foreground and background. The one thing I ran into on all three of the devices I tried it on was over heating issues. Once the phone started getting taxed, it would over heat and shut down. Literally, it would be extremely hot to the touch.

So, if I was in my regular commute and had the app running to record my drive, as well as maps (google maps or waze) running, and maybe Pocketcasts or iTunes or Play Music, it would overheat and shut down the phone.

It wasn't much use on iOS since I couldn't see the map while I was driving, and on Android, I couldn't use it in conjunction with maps and be able to listen to audio and have maps because of heating. In short, I had to make compromises in order to use it. I like the idea of it, and my usage of it by itself was really nice, but once I started adding to the mix, or wanted to make it part of my regular routine, I hit road blocks.
M
eran shirΒ β€” Co-founder & CEO, Nexar
@thehashrocket Hey Jason, thanks for the thorough testing! Android's version is still in beta and we expect the overheating issue to be resolved in the coming weeks. As for iOS and navigation, you are very right that there's a gap there, and while we do have an initial integration with Waze, this is just a first step. Again, thanks for the enthusiasm and kind words, and do check us back from time to time as we improve.
Jason ShultzΒ β€” Senior Software Engineer, 3form
@eranshir thank you for the response, I appreciate it. Perhaps I wasn't clear in my initial comment. It overheated in all three devices. Not just Android.
M
eran shirΒ β€” Co-founder & CEO, Nexar
@thehashrocket That's really weird. If you're up for it, please email me at eran@getnexar.com and I'll connect my team so we can investigate further.
Thanks!
Joshua TalleyΒ β€” Product Support Team Lead, Grundfos
@thehashrocket Great synopsis. I didn't use it on iOS because of storage limitations. It worked fairly well on my Pixel, aside from some regular initial crashes (feedback sent). I'm pretty sure this app combined with a couple others is responsible for the demise of my first Pixel (it froze and would not restart). Made me reluctant to install and use again on the replacement. Still love the idea - it's the best and only one I've seen and liked.
Commenting is limited to those invited by others in the community.
Login to continue.