Dash - a new website to view and analyse dash cam videos

[edvision]

Hello everyone and happy new year!

After getting a dash cam a few months back (Viofo A119 V3 - thanks for the recommendations!) and talking to some friends with dash cams, I felt there was a desire for a dash cam centric web platform. So I decided with some friends to build a web application to store, organise and analyse dash cam videos. We do not only show your videos, but also GPS, speed, and additional information extracted from video files.
The web app will be live in early February but in the meantime I wanted to share some screenshots and also learn more from this community on what is really needed for our product to be useful to you all.

Our product is simply called Dash and our landing page is available on: https://dash.envsion.io/ . You can register interest on the landing page and we will notify you once the application is live. In the meantime, I'd love to hear from you on what you think Dash should be. Feel free to respond to this thread or PM me.

https://flic.kr/p/2ibQ9Ut

Looking forward to your feedback!

Cheers,
Ed.

 

[jokiin]

this may be useful for people that want to share a video with insurance/police etc with the addition of the map detail, speed etc

 

[edvision]

this may be useful for people that want to share a video with insurance/police etc with the addition of the map detail, speed etc

We thought enabling the user to provide the data to insurance could be useful. We haven't discussed yet with insurance companies but I hope they will accept the data from our platform.
Do you know if most insurance companies normally accept raw footage only?

 

[Nigel]

a web application to store, organise and analyse dash cam videos

Need more information,

• how many TeraBytes storage per user?
• what does "organise" mean?
• what analysis features will it have, eg motion deblur and zoom in on license plates, door ding sound audio search?

Do you know if most insurance companies normally accept raw footage only?

Depends on the insurance company, they all have different maximum levels of technology from snail mail to Youtube, but being able to cut a few small sections out and compress them to a small sharable file where they can get the idea of what happened in a 10 second watch would be sensible, they don't like to spend much time on anything.

Clips for the police are different, they will want a couple of minutes before and after and really good quality using a standard codec and with a raw version available on request.

 

[jokiin]

I don't know that there is any standard among insurers as to what they want, a lot of luddites in the industry though so something where they could just go to a website to view could be beneficial, you're never going to get them downloading and installing a playback program

one of your challenges though is there are hundreds of different cameras using different formats for their GPS data and you need to be able to identify them and handle the files accordingly

 

[edvision]

Need more information,

• how many TeraBytes storage per user?
• what does "organise" mean?
• what analysis features will it have, eg motion deblur and zoom in on license plates, door ding sound audio search?

Thanks for the feedback Nigel.

At the moment we are still working out the user quota, whether it is in terabytes, or minutes, etc. To begin though we wanted to enable people to store many weeks of dash cam videos on the platform and learn from them.
What we mean by "organise" is:
- you can remove videos from the platform if you see fit
- you can add tags/context to videos and driving journeys
- you can share those videos

For the analysis, we are starting with traditional telematics analysis since we have GPS, speed etc (so looking at things like harsh acceleration, braking, collision, etc.) but we are going to introduce some computer vision in a few months to provide more insights. Number plate recognition would be very useful especially in crashes (particularly hit-and-run types I bet).

Can you tell me more about the audio search? What is it normally used for? I tend to mute my dash cam when I record so I am probably missing out on something...

 

[edvision]

one of your challenges though is there are hundreds of different cameras using different formats for their GPS data and you need to be able to identify them and handle the files accordingly

True - we are struggling to extract the GPS & speed data from videos taken from a few dash cams because there is no standard! But I think we are making good progress. Right now we can support Viofo A119 V3 and a NextBase one too.
Baby steps

 

[jokiin]

you won't need to offer terrabytes of data per user to still be useful, a few gigabytes might be needed as a minimum though, the minutes quickly turn into gigabytes with some cameras

Number plate recognition would be very useful especially in crashes (particularly hit-and-run types I bet).

you might find this interesting

How I replicated an $86 million project in 57 lines of code

By Tait Brown When an experiment with existing open source technology does a “good enough” job The Victoria Police are the primary law enforcement agency of Victoria, Australia. With over 16,000 vehicles stolen in Victoria this past year — at a cost...

www.freecodecamp.org

 

[Nigel]

you won't need to offer terrabytes of data per user to still be useful,

If they are offering storage then they will, a Viofo A129 Pro dual 4K can easily get through a TB per week, maybe 2 if it is in a taxi.
But if it is just for analysis then yes, upload a few minutes, do something useful with it, and delete, doesn't need much.

To begin though we wanted to enable people to store many weeks of dash cam videos on the platform and learn from them.

Remember about the dual (front+rear) and triple cameras (front+rear+interior for taxi use).

For the analysis, we are starting with traditional telematics analysis since we have GPS, speed etc (so looking at things like harsh acceleration, braking, collision, etc.)

GPS isn't great for that, normally it is a 1 second update and heavily low pass filtered so acceleration is never harsh!
The g-sensor data should be better, but tends to either not be stored or not be very useful.
Visual speed determination would make it very useful to some people.

Can you tell me more about the audio search? What is it normally used for?

When parked in the supermarket or works car park and the person in the next parking bay opens their door into yours leaving a dent, it is very difficult to find the event visually, and there generally isn't enough g-sensor data, so the way to find it is to search through all the audio to find a "ding" noise, very tedious if you do it manually! (doesn't actually sound like the word "ding", but there will be a distinct noise, followed a minute later by someone driving away and possibly preceded by a person walking up to the car.) Only works if you have a full recording with audio, such as the low bitrate parking mode on the A119 V3/A129.

 

[jokiin]

If they are offering storage then they will, a Viofo A129 Pro dual 4K can easily get through a TB per week, maybe 2 if it is in a taxi.

people that wanted to upload multiple TB of files per week would be in the extreme minority

 

[Nigel]

people that wanted to upload multiple TB of files per week would be in the extreme minority

People who want to upload anything every week will be in a very small minority!

But some people do keep asking to be able to do so.

If their website provided a standard cloud interface so that the dash cameras / phone apps can upload footage automatically, either all footage, or all locked footage, then the dashcam manufacturers such as yourself could integrate their dashcams with it and it could become very useful.

Not so easy with an A119 V3 that doesn't have wifi, but when their baby becomes a toddler...

 

[joe384]

When parked in the supermarket or works car park and the person in the next parking bay opens their door into yours leaving a dent, it is very difficult to find the event visually, and there generally isn't enough g-sensor data, so the way to find it is to search through all the audio to find a "ding" noise, very tedious if you do it manually! (doesn't actually sound like the word "ding", but there will be a distinct noise, followed a minute later by someone driving away and possibly preceded by a person walking up to the car.) Only works if you have a full recording with audio, such as the low bitrate parking mode on the A119 V3/A129.

You mean like this:

Searching parking mode footage

Came back to my car to find someone/something has dinged the door quite badly. As some of you in the A129 forum know, I've been having issues with my camera so I've had the rear camera disconnected for a while. I do however have the 8 hours of my car parking (low bit rate) from the front...

dashcamtalk.com

 

[edvision]

When parked in the supermarket or works car park and the person in the next parking bay opens their door into yours leaving a dent, it is very difficult to find the event visually, and there generally isn't enough g-sensor data, so the way to find it is to search through all the audio to find a "ding" noise, very tedious if you do it manually! (doesn't actually sound like the word "ding", but there will be a distinct noise, followed a minute later by someone driving away and possibly preceded by a person walking up to the car.) Only works if you have a full recording with audio, such as the low bitrate parking mode on the A119 V3/A129.

That's super interesting to know. I initial assumption was that sound on dash cams is generally not that useful in the case of analysis as most of the time it would be background music or people swearing... Detecting impact on one's vehicle via sound analysis when it is parked for instance is a very smart idea. Thanks!

 

[edvision]

Thanks for the link Joe384.

people that wanted to upload multiple TB of files per week would be in the extreme minority

What do you think would be reasonable in terms of storage offered to users? I presume most people are only interested in some footage that contain some event/incident and very few want to upload every minute of their driving.

 

[jokiin]

What do you think would be reasonable in terms of storage offered to users?

an event can easily end up being a few gigabytes with some cameras so 5gb to 10gb is probably enough of an allowance to cover what someone might reasonably need

I presume most people are only interested in some footage that contain some event/incident

agree

and very few want to upload every minute of their driving.

there are people around that want to archive all their footage for whatever reason, strongly suspect those types of people aren't going to want to do that to a third party anyway

 

[Nigel]

That's super interesting to know. I initial assumption was that sound on dash cams is generally not that useful in the case of analysis as most of the time it would be background music or people swearing... Detecting impact on one's vehicle via sound analysis when it is parked for instance is a very smart idea. Thanks!

Even while driving, sound is important. If someone knocks your door mirror off, it will not be recorded by a single or dual channel dashcam unless you have sound, even most triple channel dashcams do not capture the door mirrors on image. If they claim it wasn't them then you have no evidence without the sound. The swearing during an incident is also a good indication to the insurance/police/court that something happened and how serious it was, it is not a problem if it is impolite!

there are people around that want to archive all their footage for whatever reason, strongly suspect those types of people aren't going to want to do that to a third party anyway

There are a surprising number of people who expect a cloud connected dashcam to upload everything so that if they find a door ding a week after it happened they can pull up the footage on their phone even if they are nowhere near the car, and if they have their car stolen then the footage is safely stored in the cloud even if the thieves, or police in the case of USA, steal the dashcam. The practicalities of doing it are not something they think of, but they are the sort of people that might consider using a website like this. Most people are happy to have the footage stored on SD card, but they are the sort of people unlikely to use this website, unless it can offer them something they can't get elsewhere, such as number plate enhancement and motion blur removal.

 

[jokiin]

generally find the people that want everything uploaded to the cloud have zero idea about just how much data is involved

 

[Nigel]

generally find the people that want everything uploaded to the cloud have zero idea about just how much data is involved

They are only interested in the video, not the data, most of them will have turned the GPS data off so that the police can't proscecute them for speeding

 

[jokiin]

I was referring to file size, data transfer, not the data in the video

 

[joe384]

Depends on the insurance company, they all have different maximum levels of technology from snail mail to Youtube, but being able to cut a few small sections out and compress them to a small sharable file where they can get the idea of what happened in a 10 second watch would be sensible, they don't like to spend much time on anything.

Clips for the police are different, they will want a couple of minutes before and after and really good quality using a standard codec and with a raw version available on request.

A lot of police forces in the UK allow you to upload dashcam footage, apparently though GPS from a dash cam in not really worth anything in court.

Some insurance companies do have websites where you can upload footage, for some of the small companies that don't have this it might be very benificial to get the customer to use a site like this though, and being able to include g-sensor and gps data is a great advantage.