Challenge
Outrider is the leader in autonomous yard operations for freight movement.
Outrider started off with 2D images for their autonomous trucks. However, 2D lacks the depth and dimensions crucial for recognizing and processing the varied trailers in a supply chain yard. When improving their perception models from 2D images to 3D scene fusion, the engineering team needed labeling capabilities that would go beyond their current labeling platform, which was restricted to 2D.
A best-in-class labeling platform had to solve specific issues next to the standard AV challenges like people, trucks, …. For example, reading the ID on trailers, pinpointing connection points, and ensuring uniformity in trailer dimensions.
Their setup with 10 cameras and 4 LiDARs necessitated a solution that could seamlessly merge diverse data points.
Solution
With Segments.ai, the perception team unified all labeling projects under one roof, ensuring streamlined operations.
With the multi-sensor capabilities, they further accelerate the labeling process by projecting labels from 3D to 2D scenes. Next to speed, this allows the labeling team to link 2D and 3D data labels consistently and without fault across time and modalities.
The labeling service allows Outrider to scale the workforce swiftly, both for labeling and reviewing annotations, to meet the rising demands of their expanding operations.
Benefits
Get multi-sensor label projections
Label objects in the 3D point cloud interface first. And project your 3D cuboids to 2D bounding boxes on all camera images.
Get consistent tracking IDs
Get a consistent track ID for an object over multiple frames and across various sensors
Cross-sensor visualizations
By overlaying sensors, you provide your labeling team with more context, which leads to more accurate labeling and fewer errors. Ultimately saving you time and money.
