Evotegra: Unleash roadmarkings to EVO/2MP/TRFC/DE/200K

Name of project:

Unleash roadmarkings to EVO/2MP/TRFC/DE/200K dataset

Short description:

Extend the EVO/2MP/TRFC/DE/200K Dataset with new road marking classes like dashed, wide dashed, solid, double solid or arrows

Long description

Currently the EVO/2MP/TRFC/DE/200K dataset contains 192,499 images with 807,415 individual labels in 280 classes. With this proposal we are going to extend the dataset by at least the following road markings:

  • dashed lane
  • wide dashed lane
  • solid lane
  • double solid lane
  • arrows
  • zebra crossing

This probably will extend the dataset to over 2 million unique labels and further improve the value of the dataset. In our highly automated process all labels are created and verified multiple times by an AI supervised by humans. The improved quality and consistency of our data enables AI to unlock its full potential.

How does this project drive value to the Ocean ecosystem?

Our goal is to create the light house project for AI image processing data on the Ocean Market, attracting consumers to the market and help to establish Ocean Markets as a prime source for high quality data.

Category

Unleash data

Proposal Wallet Address

0x61B15998893cC746B46C08FEdEE13a0d1b33bBa9

Team information

Website: www.evotegra.de
Contact Email: manthey@evotegra.de
Twitter Handle: @evotegra
Discord Handle: Tobias Manthey#8624
Country of residence: Germany

Funding Amount:

16.500 Ocean

Current Remaining Grant Treasury Balance

0

Have you previously received an OceanDAO Grant?

Yes

2 Likes

Current status: work in progress

Current status: Completed

We delivered 5 of 6 classes. During the project it turned out that wide dashed lanes lacked available data and separation from normal dashed lanes. To compensate the missing class, we delivered another 132.000 annotations (26%) over the promised minimum amount.

We updated the EVO/2MP/TRFC/DE/200K dataset with the classes:

[X] dashed lane
[ ] wide dashed lane
[X] solid lane
[X] double solid lane
[X] arrows
[X] zebra crossing

with at least 500000 new AI validates instances.
[X] 632415

extend the dataset to over 2 million unique labels
[X] 2035145