# Journey into ROVR - Go Paris Edition

Hello everyone, welcome to the latest article in our *Journey into ROVR* series—*Go Paris Edition*. This article will provide a detailed update on the latest developments of our ***Go Paris*** project.

## **First, what is Go Paris?**

We’ve never provided a detailed introduction to the background of the ***Go Paris*** initiative, so in this article, we’ll take the opportunity to share the origins and progress of the ***Go Paris*** project.

As many of you know, ROVR is committed to the collection and generation of high-precision 3D data. With our two dedicated hardware devices, the TarantulaX and LightCone, we can collect and generate precise 3D data. One of the key products of this data is HD Mapping, used in autonomous driving.

The ***Go Paris*** initiative originated from a Proof of Concept (PoC) project with a European customer. Through urban-level HD Mapping (which in the web2 space is referred to as purely visual crowdsourced HD Mapping), we validated ROVR’s mapping capabilities, as well as the mapping coverage and frequency within the ROVR community. (Due to confidentiality agreements, we are currently unable to disclose details of the partner. We will negotiate further disclosure of relevant information and vehicle models with our partner after the PoC is completed.)

The PoC is divided into three main stages:

1. **PoC Area Selection:** This was determined by our partner. The selection criteria were large metropolitan areas within Schengen countries, with relatively low web3 activity. Paris was chosen as the city for this PoC, to validate ROVR’s large-scale mapping capabilities and the coverage and frequency of road mapping in areas with low web3 activity.
2. **HD Mapping:** This is handled by ROVR. ROVR is responsible for mapping the selected area, and we provide data submissions to our partner weekly. The project has now entered its seventh week.
3. **Accuracy Validation:** This is managed by our partner, with ROVR collaborating. The partner will evaluate the accuracy of the location data and assess road element accuracy and recall rates for the data provided by ROVR.

## **Second, how is Go Paris progressing?**

The ***Go Paris*** project began on November 18, 2024, and has now been running for 7 weeks. Currently, the project is in the second phase—HD Mapping. On January 6, 2025, ROVR submitted the seventh data set to our partner.

Over the course of 7 weeks, with help from the ROVR community, we have mapped 126 square kilometers of the core area of Paris (10 x 12.6 km), with a total accumulated distance of over 1,500 km.

Below are the statistical results of our HD Mapping efforts (as of the seventh submission on January 5, 2025):

* **Total Distance Mapped:** 1,500 km+
* **Traffic Signs:** 36,930
* **Road Markings:** 49,956
* **Crosswalks:** 15,653
* **Lane Markings:** 208,780 segments, covering over 6,000 km+

<figure><img src="/files/xjziNCpvmVgoXsGroz5v" alt=""><figcaption></figcaption></figure>

Together with our partner, we conducted accuracy evaluations on certain features(the number of evaluations is limited, and the metrics are for reference only). The results are as follows:

|               | Accuracy Rate | Recall Rate | Avg. Position Accuracy |
| ------------- | ------------- | ----------- | ---------------------- |
| Traffic Signs | 87.0%         | 91.7%       | 0.112m                 |
| Road Markings | 92.5%         | 89.3%       | 0.191m                 |
| Lane Markings | 90.7%         | 89.3%       | 0.147m                 |

In the web2 industry, the accuracy requirements for crowdsourced HD maps are typically a position accuracy within 20 cm and a precision/recall rate of over 85%. Due to insufficient revisit frequency on some roads, we will continue to update the data.

In areas with more than 40 revisits, the accuracy and recall rates of the three elements—Traffic Signs, Road Markings, and Lane Markings—have met the precision requirements of the HD Map(P/R>85%, Abs. err < 0.2m).

For this accuracy validation, we used data from traditional HD Map providers as ground truth and conducted a sampling verification.

However, we believe that **without an evaluation method, any precision is just a number**. Therefore, we will publish the evaluation method.

**The evaluation method is as follows:**

**Definition:**

**Set R**: The set of evaluation units formed by all reported linear features.

**Set T**: The set of evaluation units whose absolute accuracy satisfies the specified threshold

**Direction Definitions:**

**X direction**: The lateral direction, perpendicular to the road's travel direction in the horizontal plane.

**Y direction**: The longitudinal direction, aligned with the direction of travel of the road

**Z direction**: The height direction, perpendicular to the road's travel direction in the elevation plane.

* For discrete elements，such as Traffic Signs, Road Markings:
  * The evaluation method for the abs. accuracy of discrete elements

    d=√(((𝑥′−𝑥)^2+(𝑦′−𝑦)^2+(𝑧′−𝑧)^2)/2)
  * Abs. Accuracy Compliance Rate Calculation:

    d\_j=√(((x\_j′−x\_j )^2+(y\_j′−y\_j )^2+(z\_j′−z\_j )^2)/2)
  * Compliance Rate:

    P=C\_t⁄C
* For continuous elements，such as Lane Markings:
  * The evaluation method for the abs. accuracy of continuous elements

    di=√(((𝑥i′−𝑥i)^2+(𝑦i′−𝑦i)^2+(𝑧i′−𝑧i)^2)/2)
  * Abs. Accuracy Compliance Rate Calculation:

    d\_j=√(((x\_j′−x\_j )^2+(y\_j′−y\_j )^2+(z\_j′−z\_j )^2)/2)
  * Compliance Rate:

    P=L\_t⁄L

## **Third, will the Go Paris project continue in the future?**

At present, the feedback from our partner has been very positive, recognizing both the activity level of the ROVR community and ROVR’s technical capabilities. The ***Go Paris*** project will continue, with ROVR continuing to optimize the HD map data and submitting weekly updates.

The area we are showcasing this time is part of the HD Map generated by the Go Paris project. We will continue to expand and complete the full HD Map mapping of Paris in the future.

Within the scope of the data presented this time, there are still some road segments that have not been covered (yellow arrows) or have insufficient revisit frequency (blue arrows), as shown in the image below. We will continue to increase the coverage and frequency of data collection in Paris in the future.

<figure><img src="/files/gGtdu0xJKbekX1FkotlK" alt=""><figcaption></figcaption></figure>

We will also provide **a&#x20;*****web-based visualization platform(***[***https://rovr.network/map/***](https://rovr.network/map/)***)***, allowing the community and other data users to gain a more intuitive understanding of the ***Go Paris*** project's progress.

In the future, we plan to make the HD map data generated by the ***Go Paris*** project publicly available for free.

## **Conclusion**

Map serves as a digital representation of the real-world transportation infrastructure, capturing the detailed features of road assets. Mapping, therefore, goes beyond simply uploading images—it's about extracting high-value data that underpins informed decisions in infrastructure development and management.

Thank you very much for taking the time to read through our article. **We eagerly look forward to your participation in accelerating ROVR's growth and achieving greatness together.**

For more information, please visit our official media channels:

🔸 ROVR Network ([\*\*\*](https://rovrlabs.io/)<https://rovr.network/#/home***>)&#x20;

💬 X: ROVR ([\*\*\*](https://x.com/ROVR71776)<https://x.com/ROVR_Network>[\*\*\*](https://x.com/ROVR71776***))&#x20;

🗯 Discord ([\*\*\*](https://discord.com/invite/RjV3E3u4F2)<https://discord.com/invite/RjV3E3u4F2***>)

Thank you!

ROVR Team


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://rovr-network.gitbook.io/rovr-docs/journey-into-rovr/journey-into-rovr-go-paris-edition.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.