3D cuboid labeling remains one of the more time and cost-intensive steps in building perception datasets. At Latitude, we’ve developed an in-house auto-labeling pipeline to seed our human labeling workflows, significantly reducing labeling time and cost. In these videos, you can see auto-labels generated by our in-house pipeline on unseen scenarios for an entirely new vehicle platform. The pipeline comprises of two stages. First, a bird’s eye view detector fuses data across time and multiple sensors (e.g., cameras, LiDAR and RADAR), and it is paired with a learned tracker. The detector is designed to be configurable with respect to the type and number of sensors used as input, so that it can leverage the full sensor suite available on different vehicle platforms. Second, tracks are refined separately for static and dynamic objects using learned LiDAR-based refiners and clustering. This refinement is key to producing auto-labeled tracks with high precision and accurate kinematics across diverse scenarios - from high-speed highways to low-speed parking scenarios with many parked vehicles. Our auto-labeler lets us set a high quality baseline without compromising on volume, while retaining the ability to provide human correction to a subset of the most valuable examples. We're constantly iterating on how to find the right balance between efficiency and fidelity. If you're interested in working on real-world perception problems, including: dataset tooling, scalable labeling pipelines, and applied machine learning then check out our open job listings: 🔗 https://lnkd.in/eXpXtm7g #MachineLearning #ComputerVision #3DPerception #DataLabeling #Autonomy #AutoLabeling #Engineering #Hiring
Latitude AI
Technology, Information and Internet
Pittsburgh, PA 14,506 followers
Reimagining what it’s like to drive.
About us
Latitude AI is a wholly owned subsidiary of Ford Motor Company that develops automated driving technologies. The company was formed in 2023 to reimagine what it’s like to drive, giving drivers time back on their journeys and transforming the driving experience to be safer, less stressful, and more enjoyable for everyone. Headquartered in Pittsburgh with engineering centers in Dearborn, Mich. and Palo Alto, Calif., Latitude employs about 550 people.
- Website
-
https://lat.ai/
External link for Latitude AI
- Industry
- Technology, Information and Internet
- Company size
- 501-1,000 employees
- Headquarters
- Pittsburgh, PA
- Type
- Privately Held
Locations
-
Primary
Get directions
2545 Railroad Street
Pittsburgh, PA, US
-
Get directions
17000 Federal Drive
Dearborn, MI, US
-
Get directions
3251 Hillview Ave
Palo Alto, CA, US
-
Get directions
25 Steel City Street
Greenville, SC, US
Employees at Latitude AI
Updates
-
Do you want to work with the team making a real, positive impact on the driving experience for millions of people? We are hiring senior engineers to join our team. If you want to meet our team and learn more about our open roles, we are onsite at #CVPR through Sunday with senior members of our engineering team. Stop by our booth to learn more about our development of automated driving technology at scale. Learn more about Latitude at lat.ai and apply here: https://lat.ai/careers #hiring #technologyjobs #automateddriving #engineeringjobs #CVPR2025
-
-
Live from #CVPR2025! Latitude is here in Nashville and our team is excited to be at CVPR. Come chat with us, see our test vehicle, and learn more about the company at Booth #923. #computervision #automateddriving #nashville #CVPR2025
-
-
Our team at Latitude AI is working on making a real, positive impact on the driving experience for millions of people. We are on a mission to give drivers time back on their journeys and transform travel to be safer, less stressful, and more enjoyable for everyone by developing automated driving technology for future Ford vehicles. If you’re interested in joining us on our mission, visit 👉https://lat.ai/careers