Map Improvement Proposal 17 (MIP-17)
6/14/2024: After a healthy discussion, and analysis of community feedback, we have decided to finalize the proposal with a cutoff of 65 kph for side-facing imagery, as proposed. We gave serious consideration to suggestions of a higher limit of 70 kph or 75 kph, but we did not see a strong case for why these would better support customer needs. For more details on this, see the “Input on the Proposal” section below.
This change will take effect for the week of July 8, with rewards issued on July 17, 2024. This will give contributors with side-facing dashcams a few weeks to make arrangements, such as giving a dashcam to a friend, deploying it as part of a fleet, or selling it secondhand.
A note on implementation: Trimming of side-facing imagery will be implemented at the start of network processing, rather than on the dashcam or the app. This means drivers with side-facing dashcams will be able to upload side-facing imagery taken at high speeds, but it will not be fully processed. The reason is that mount position is assessed during network processing. In order to prevent side-facing imagery from being uploaded at all, there would need to be an additional edge processing step, slowing down uploads and further straining the capabilities of HDC. It is possible that we will move mount detection to the edge on the Bee, which has more computing power to accommodate this step.
TL;DR
Since the launch of the network, we’ve learned that map customers have little demand for side-mounted imagery outside of urban areas. For this reason, we propose only accepting imagery from side-mounted dashcams below 65 kilometers per hour (approximately 40 miles per hour) to focus on denser areas. We expect this proposal to have a slightly positive impact on Global Map Progress and the contributor reward pool, while redirecting more rewards toward front-facing dashcams.
Background
Since the launch of the Hivemapper Network, there have been four approved mount positions: forward-facing interior, forward-facing exterior, side-facing interior and side-facing exterior. We have allowed a maximum of two dashcams per vehicle: one forward-facing and one side-facing.
These mount positions even predate the launch of the Hivemapper Network. For a blast from the past, check out this 2021 blog post from the early alpha and beta stages of network development, when the project team tested smartphones and off-the-shelf dashcams and found that purpose-built hardware was critical to effectively mapping the world.
This was the philosophy on front and side mounts at that time:
Front facing mounts are focused on mapping the street (traffic signs, lanes, parking, etc.) while side facing mounts do a better job of mapping ‘doors and stores’ (building addresses, entrances, store signage, etc).
It was true then, and it’s true now. However, with the rapid growth of the Hivemapper Network, we’ve learned that network policies are misaligned with this philosophy and the needs of map data customers.
Customers are indeed VERY interested in sidewalk-facing map data in urban areas, where it is useful for “doors and stores” data points. However, it is now clear that customers are NOT interested in side imagery on rural roads and highways — draining resources from the project in multiple ways.
In fact, major map data customers have been actively filtering out the side-facing imagery that they do not wish to buy in these areas.
Proposal
To focus on “doors and stores” as intended since before the launch of the network, we propose a change to the network’s approved mount positions so that side imagery will only be accepted when traveling less than 65 kilometers per hour, or approximately 40 miles per hour.
We are proposing this limit because we believe it will ensure side-mount imagery is almost always accepted while driving safely in dense urban environments where “doors and stores” data points are valuable, and almost always rejected while traveling at higher speeds on rural roads or highways. There will inevitably be some false negatives and false positives, but we believe 65 kilometers per hour will best minimize these.
We are open to input on the numerical limit and the exact implementation approach, as long as it is aligned with the needs of the network.
The success of the network relies on efficiently building products that customers want — and aligning incentives to fairly reward the contributors who are making that possible.
This change would require technical implementation. In addition, it is only fair to provide advance notice to allow people to redeploy side-mounted cameras. For this reason, if this proposal is finalized, we expect it to be implemented no sooner than July 1, 2024, and perhaps a bit later.
Impact on Global Map Progress
Contributors may have questions about the impact of this proposed change on Global Map Progress and the rewards pool. While the impact is difficult to precisely project, we expect a neutral to slightly positive impact on Global Map Progress and the weekly rewards pool.
To be more specific, we expect a small positive impact on Coverage, which accounts for 60% of Global Map Progress. We expect this to outweigh a small negative impact on Activity, which accounts for 30% of Global Map Progress. Resilience, which accounts for 10% of Global Map Progress, should not experience a meaningful change if this proposal is implemented.
Coverage: Front-mounted and side-mounted dashcams usually travel within the same hex. Although freshness is calculated separately, the Coverage metric does not account for mount position when calculating whether a hex has been covered. This means that driving with a second dashcam has no added benefit for Coverage. It also means that if some side-mounted dashcams go idle after this change is finalized, that would have no impact on Coverage. However, if these side-mounted dashcams are redeployed as front-mounted dashcams in separate vehicles, that would have a positive impact on Coverage. On balance, we expect a modest positive impact on Coverage.
Activity: Front-mounted and side-mounted dashcams both contribute to Activity, which is a measure of the number of total road km submitted in a region. If some side-mounted dashcams go idle after this change is finalized, that would have a negative impact on Activity. If those side-mounted dashcams are redeployed in new vehicles, that would have a neutral impact on Activity. On balance, we expect a modest negative impact on Activity.
Resilience: Front-mounted and side-mounted dashcams both contribute to Resilience if managed using distinct wallets. A side-mounted dashcam does not contribute to Resilience if it is managed using the same wallet as a front-mounted dashcam. If side-mounted dashcams that were previously managed using distinct wallets go idle, it would have a negative impact on Resilience. If side-mounted dashcams that were previously managed using the same wallet as a front-mounted dashcam are redeployed as front-mounted dashcams in new vehicles, that would have a positive impact on Resilience. It is unclear which scenario will be more common. On balance, we expect a neutral impact on Resilience and Global Map Progress.
There are many unknowns that make it difficult to project the exact impact. Most notably, we do not know what share of dashcams would be idled versus redeployed. However, considering that Coverage is the largest factor in Global Map Progress, we are confident this change should result in a positive impact on the size of the rewards pool.
Frequently Asked Questions
- Why do map data customers prefer front-mounted dashcams?
The Hivemapper Network is focused on street-level navigation maps. Most of the objects that comprise a navigational map are designed to be visible from the perspective of a driver, putting them in the field of view of a forward-facing camera. Most current customers use Hivemapper Network data for navigational use cases, where road features are most important.
Some points of interest are more visible from a side view, especially when intended for a pedestrian on a sidewalk. In industry slang, these are called “doors and stores.” These points of interest are overwhelmingly found in urban areas, rather than rural roads or highways. Customers have conveyed they do not see meaningful value in side-view imagery on highways. Some of them have actively filtered out side-view imagery when consuming network data.
2. Why not just keep rewarding side-mounted dashcams everywhere? Maybe a customer will want that imagery later.
It’s possible, but we think it’s unlikely based on feedback from customers.
If we want the Hivemapper Network to serve customers as efficiently as possible to support our mission of building the world’s freshest map, it is critical that incentives reflect the value being created. Issuing rewards for map data that is unlikely to be consumed is not a good use of the limited resources available in the rewards pool.
3. What about 360-degree imagery?
It might make sense to incentivize 360-degree imagery in the future to compete toe-to-toe with consumer-oriented offerings such as Google Street View, but that is not currently a priority compared to front-facing stereo cameras with edge processing, like the Bee.
4. Why not geofence urban areas where side-mounted imagery is accepted?
We are building a decentralized, permissionless mapping network. Unless there are very compelling reasons to do so, we must avoid choosing specific areas in the world where people can map. The speed limit approach is regionally agnostic. Instead, it is based purely on objective math, lining up with common patterns of automobile-oriented urbanization.
5. Why did you choose 65 kilometers per hour?
Most countries set speed limits between 30 km/h and 60 km/h for urban roads, with higher speed limits for rural roads and highways. (For a reference, see the World Health Organization.)
We believe a limit of 65 kilometers per hour (40 miles per hour) will ensure side-mount imagery is almost always accepted while safely driving in the dense urban environments where side-facing imagery is most valuable to customers, and almost always not accepted at higher speeds on rural roads or highways. There will inevitably be false negatives and false positives, but we believe 65 kilometers per hour will best minimize these.
Although the numerical standard means drivers would occasionally be able to submit side-mounted imagery on highways (for example, during a traffic jam) we would expect most drivers to self-select based on the types of roads they typically drive.
For example, a truck driver who mostly drives on the highway might choose to drive with only one camera, while a rideshare driver who mostly drives in the city center might choose to drive with two. Every driver should make their own decisions about what makes sense for them. However, we think most highway drivers using two dashcams would generate more useful data (and more rewards) by allocating their second dashcam to a fleet and splitting rewards with a friend who mounts it forward-facing.
We also strongly discourage anyone from intentionally driving slower than the threshold on roads where it is inappropriate to do so. Mapping should be passive based on the driving you are already doing in your day-to-day life, and should never take precedence over driving safely.
6. I mostly drive on the highway with two dashcams. What are my options?
We recognize that some people have grown accustomed to driving with two dashcams and will be disappointed by this proposal. We do not take this decision lightly and we hope you understand why the network is best served by aligning rewards with customer needs.
To be clear, contributors using a side-mounted dashcam on the highway did nothing wrong. They followed the rules and did what seemed to be best at the time. Many of our most devoted contributors fall into this group.
In retrospect, supporting side-facing imagery in 100% of circumstances was the wrong choice. We believe the long-term health and success of the Hivemapper Network demands a mindset of rapid iteration to serve customers, even when it is frustrating in the short term.
If you are mostly driving on high-speed roads with two dashcams, you have a few options:
- Give one dashcam to a friend or relative
- Start a fleet and receive a token split
- Sell your dashcam secondhand (for a fair price, please!)
- Keep driving with two dashcams and only map urban areas
7. Why have you chosen an effective date of July 1, 2024?
We usually move quickly to implement changes that will improve the health and potential of the Hivemapper Network.
Although this change could be implemented fairly quickly, we would continue to reward side-mounted camera imagery without restriction until the effective date of no sooner than July 1, 2024 to ensure all contributors have time to choose and implement one of the options above, even if that involves giving a dashcam to a friend, relative or fleet member.
8. Will this change also apply to the Bee?
We have previously communicated that we expect to implement changes to rewards pool formulas after the launch of the Bee to account for dashcams that can generate map features on the edge.
This could have implications for our policies related to side-mounted dashcams. On the one hand, most customers would be agnostic about the mount position that generated object detections. On the other hand, detecting the same map feature twice at the same moment in time (once with a front-facing camera and once with a side-facing camera) is not worth twice as much as detecting the map feature once. There would need to be a freshness component to map features, as with imagery, which would naturally reduce the rewards issued for capturing the same map feature with multiple cameras.
Input on the proposal
Active members of the Hivemapper community were encouraged to participate by sharing questions, suggestions, and comments in the #mip-17 channel of Hivemapper’s Discord server. The scheduled comment period was 7 days, running from Monday, June 3, 2024 through Monday, June 10, 2024, but we continued to review input through Friday, June 14.
Some contributors suggested raising the threshold from 65 kph in the proposal to 70 kph, 75 kph or 80 kph.
“In the city, especially when passing people, it’s not uncommon to get above 65.”
“There are slower highways filled with commercial/residential zones and the speed limit on those roads are [65–72 kph].”
“In the area where I live, there are loads of businesses on main roads that have a speed limit of [64 kph] (occasionally even [72 kph]). The highway speed limit is [89 kph], so even a limit of 80kph would still easily exclude the highway, especially since the actual flow of traffic is usually [105–113kph] or more.”
We did not consider 80 kph. Although this is lower than the highway speed limit in many countries, including the United States, it is a common highway speed limit in many countries.
We gave serious consideration to the suggestions of 70 kph and 75 kph. Although we have not seen meaningful customer interest in side-mounted imagery from “stroads” with higher speed limits, we wanted to evaluate the idea that the network would lose valuable data as a result of a lower limit.
Contributors shared several examples of roads where they thought a lower speed limit could cause the rejection of valuable side-mounted imagery.
One contributor shared an example of Google Street View imagery from Linden, NJ, USA. The image, dated November 2022, showed a posted speed limit of 45 mph in front of an Appliances Outlet store.
We pulled front-facing and side-facing imagery from the Hivemapper Network collected in May 2024.
The name and phone number of the store were clearly visible from a front-facing as well as side-facing view. Both images showed that the 45 mph speed limit sign which was visible in the Google Street View image from November 2022 was not present — demonstrating the importance of fresh imagery to establish the reality of road conditions for mapmaking.
We could conceivably imagine data points that would be valuable from a side-facing collection. For example, the street address on the door is more legible in the side-facing image. However, considering the lack of customer interest in side-facing imagery from this type of location to date, it does not seem sufficient to justify issuing double rewards for a single drive and burdening the network with added processing costs.
Another contributor shared an example of a Google Street View image from Tower Ave., in Superior WI, USA, dated August 2023.
We took a look at Hivemapper Network imagery and noticed the speed limit on this road was actually 35mph, meaning that a driver traveling this road at the legally required speed would not be prevented from uploading the imagery and receiving rewards for it under the MIP-17 changes.
Regardless, we evaluated the example. It is important to remember that on a high-speed road, the “doors and stores” tend to be specifically designed to be visible to drivers — meaning that a front-facing dashcam usually collects the same information. In the specific example provided, all the business signs in the shopping center were also visible on a sequence of multiple front-facing images; with multiple “shots on goal,” a customer would receive the same information from front-facing imagery.
We did not find any recent side-facing imagery from this location. However, front facing imagery was enough to observe that the nail salon “Crystal Nails & Spa” that was visible in the Google Street View image from August 2023 had been replaced by a business called “Nail Envy.”
As with the prior example, we do not see a strong argument for side-facing imagery on these kinds of high-speed roads, when the “doors and stores” are optimized to be visible to drivers.
We recognize that changes in the rules of the network can be jarring and frustrating. Our community should not take it lightly when rule changes prompt contributors to change behavior or rethink their decisions about how to participate. But as always, we are putting customers first; efficiently building a great product that they want to buy is the only way to develop a sustainable network that accrues value to everyone involved.
Thank you to everyone who shared input and suggestions to help ensure we would arrive at the best decision possible. Our community must always move quickly to serve customers, and we should not hesitate to reverse or modify this decision in the future if it better serves the map.
