COVID19 Threat or Opportunity for transport planning companies?

The coronavirus outbreak has heavily impacted transport planning companies among the many others. The mobility within almost every city around the world changed due to the freedom of movement restrictions, reactivation of border controls, and many more restrictive provisions. It is evident that transport planning has been frozen due to the fight against the coronavirus. However, Is it an opportunity or threat for the future? What are the roles of AI and Visual analytics within the current and post-coronavirus recovery road map? What business impact can the use of AI bring to transport planners and surveyors in scaling their post-crisis business?

Q&A

Q #1:

Can separate video feeds be combined on the platform to create a network-wide origin-destination analysis or journey time analysis? (like ANPR - Automatic Number Plate Recognition)

A #1:

Currently, Video Insights allow for origin-destination analysis on the micro-level. However, the ANPR module is in the final stages of evaluation. The platform will provide the licence plate recognition for origin-destination studies on the macro level in the matter of weeks, including the journey time analysis of the vehicles. This feature will be available for vehicular traffic with licence plates. We will keep you updated about new releases. Contact us via email or via live chat to discuss the pilot assignment with this technology.

Q #2:

Do you support live camera processing or real-time processing?

A #2:

Yes, Video Insights supports both live processing of camera feeds and also the on-site real-time traffic monitoring using the appliance connected to the camera. In all cases of video provision, the extracted anonymous data will be transferred to the central Video Insights platform (your account) for more in-depth analysis.

Q #3:

Can GoodVision provide calibration parameters for systems like PTV Vissim?

A #3:

Yes, transport modellers have a wide variety of analyses available over the collected data, that help them to create and calibrate traffic models. GoodVision provides, i.e. saturation flows, free-flow intervals, travel time, headway time (time-gap), travel speed, decreased speed areas, acceleration and braking patterns. All for the traffic flow as an average of each vehicle/ped. And of course the multimodal traffic counts and peak hours.

Q #4:

Can you process night video footages?

A #4:

Yes, Video Insights AI is also trained to handle night footage. The ability to detect vehicles in the night is defined by the visibility of the object contours – meaning there has to be some illumination – from the lamps or the headlights of other vehicles. If the naked eye can recognise the object, then GoodVision can too.

Q #5:

Can you do ANPR on a wide network if you use drone video footage?

A #5:

The ANPR module is in the final stages of evaluation. Our platform will provide the licence plate recognition for origin-destination studies on the macro level in the matter of weeks, including the journey time analysis of the vehicles. This will be available for vehicular traffic with licence plates. We will keep you updated about new releases. Contact us via email or via live chat to discuss the pilot assignment with this technology.

Q #6:

I have used your platform in detecting traffic inroads. How accurate is your platform in counting people in indoors like shopping malls?

A #6:

Video Insights supports pedestrian monitoring, and it is well trained for it. Indoor tracking is also possible if the video requirements are met (read more here) Mall streets are supported. However, the close-up scenarios like in-store (boutique) monitoring are not supported out of the box.

Q #7:

Can you help traffic modellers save time with preparing the default traffic models, i.e. by calculation of saturation flows?

A #7:

Yes, see the previous answer. Saturation flows are supported, including many other traffic parameters that are helping modellers to conserve time on manual video review. I.e. Saturation Flows by default configured to support the UK TfL guidelines. Still, the app itself allows you to set it to support the directions in your own country easily.

Q #8:

Would the sat flow measurement include cyclists?

A #8:

Yes, saturation flow calculation can include cyclist if it makes sense to you in your model and if you select bicycles in your filter.

Q #10:

Does AI technology work for Indian roads, as there is no lane discipline and driver behaviour

A #10:

Yes, GoodVision Video Insights works worldwide, including Indian roads and it’s traffic specifics. No-lane or mixed-lane behaviour is not a show-stopper. GoodVision detects vehicles on their positions wherever they are, regardless of the streets or lanes, you don’t need to define where the roads and pavements are, we detect all flows automatically.

Q #11:

Are there any deployment needs? Or integration needed? Can we use our cameras? Do you also have a desktop app?

A #11:

GoodVision is a SaaS and compatible with your existing CCTV infrastructure or temporary cameras, so there are no deployment or integration costs. You can do your first project immediately upon registration. Video Insights can be accessed via the web at https://my.goodvisionlive.com or via the desktop app available here: https://my.goodvisionlive.com/download.html

Q #12:

What sort of accuracy % can you guarantee? For counts/ settings/ classifications etc?

A #12:

If the requirements for video quality are met, GoodVision will provide 95% accuracy on traffic counts and classifications. Learn about the video requirements for standard cameras and drone high-detail processing. Please do not be confused with the common saying that AI cannot be 100%. It is true. But GoodVision counting accuracy is not given just by AI accuracy. We implement plenty of other algorithms and probabilistic methods of vehicle tracking, which fill the gap between AI accuracy. In any case, we can proudly state that GoodVision’s AI is a state-of-art proprietary technology and on the absolute top in our industry, trained on one of the biggest traffic-specific datasets, particularly selected and gathered by our analysts.

Q #13:

Does AI technology classifies Indian classification

A #13:

Yes, GoodVision provides the standard classifications also on Indian roads. If your question is about the specific vehicle classes like tuk-tuks, these classes can be added to Video Insights on the individual basis. PS: tuk-tuks will be supported in GoodVision soon as standard.

Q #14:

How about the data privacy of my data? GDPR regulation

A #14:

GoodVision Video Insights is fully GDPR compliant. It extracts only the anonymous data from the video footage, the processing happens in the region of the customer, and the video files are immediately deleted after the processing, which takes 1 hour. Our on-site appliance offers full data-private processing on site. It sends only anonymous extracted data to the cloud platform. Our service is privacy shield protected, and we can provide you with this declaration in writing. We are not storing data on any private servers or datacenters.

Q #15:

Can it be used out of the box anywhere (like in US, Australia) or do you need to customise and train for that location first?

A #15:

Yes, it is prepared to work out-of-the-box anywhere in the world, and no customisation is required.

Q #16:

How long will it take to collect traffic data from ~5000 hours of video?

A #16:

One hour, 1hour +- 15 minutes exactly. Any number of video-hours uploaded at once will be processed within ~1 hour upon upload. This is thanks to our unique super-parallel processing.

Q #17:

Can the traffic data reports be exported into Excel or CSV files?

A #17:

Yes, all traffic parameters and counts can be exported into Excel reports using our report builder. Reports are generated immediately, can be created by the user anytime, repeatedly, and they are

for free. Traffic data can also be exported in the form of diagrams and visual maps directly from the app.

Q #18:

If we do have specific classifications, will it delay the turnaround time?

A #18:

No. If the system is trained for the particular class, the turnaround time will stay the same 1 hour.

Q #19:

How many classes of vehicles can you détect ? do you detect bicycles? Segway?

A #19:

GoodVision classifies vehicular attendants, pedestrians and bicycles – 8 classifications in total according to UK classification scheme and US Highways classification scheme: car, van (LGV), bus, truck (OGV1), articulated truck (OGV2), motorbike, bicycle, pedestrian.

Q #20:

Will there be an option to include the original video footage instead of the camstill in video insights in the future?

A #20:

Yes, there is such an option now – easily connect the camera feed directly to GoodVision Video Insights, and you can schedule your survey up-front, remotely from your office. GoodVision also allows you to schedule repeated surveys, i.e. daily from 8:00 am to 8:00 pm, etc.

Q #21:

Can you do a predictive analysis?

A #21:

Predictive analytics and future insights from the historical data is on our current development roadmap. We will be happy to hear your ideas and requirements on this topic to develop the feature that is helpful to you. We will contact you via the email provided.

Q #22:

Do you detect a passenger in the vehicle? How many are they? The colour of the vehicle?

A #22:

Our services are not focused on identifying passengers in cars. Yes, we detect the colour of the vehicle, and you can filter your analyses by colour.

Q #23:

Can I run a complete project on your platform without switching back and forth between systems?

A #23:

Yes, this is our intention to boost your productivity by allowing you to do all the tasks in the same app. That means you can upload the video footage (or live cams) to the app, get it processed, store the video content in our new GoodVision Vault storage, perform the analyses, counting, report the data, share the data with your team of hand-off your project to your customer (including the video files), perform the analyses of model calibration parameters, export the data in the way that simulation systems can feed on it, analyse events, review the events in the original footage in the app, etc. As you see, we are supporting the whole chain from video collection to traffic model creation without the need to change the app.

Do you have some more questions or feedback regarding this webinar? Let us know! We would love to hear them! 📩

Did this answer your question?