The CVEDIA team supports car manufacturing partners in efforts to develop ADAS/AD systems. CVEDIA is an industry leader in the use of custom synthetic data to accelerate development of autonomous vehicles. Our systems have been precisely built to minimize the gap between real and synthetic data for fluid algorithm training.

CVEDIA simulation tools are designed from the ground up to support and accelerate development from the earliest to the latest stages in the development process. Our SynCity simulation platform provides broadly representative data for new projects, and is used in data intensive classifier training to augment real world data, fill in gaps in coverage, and control for sampling bias. The result is rapidly produced, higher quality training sets that dramatically reduce field data collection, data storage, and labelling expenses.

Use Case

Ask a question

An ADAS manufacturer came to us after a few issues presented in training their algorithms. For one, their system was having a difficult time perceiving and classifying occluded objects, such as fire hydrants as people walked by. Second, creating semantic segmentation of LiDAR data was both exceedingly difficult and cost prohibitive.

Finally, a (very common) issue with testing their AI was that they could not gather effective data for the most dangerous edge cases – scenarios like crashes, children running in front of vehicles, or snow blindness. This blind spot made ensuring system safety a gamble.

CVEDIA was able to create a environment in SynCity for this client that included custom objects, such as humans and city infrastructure, and edge case weather conditions that could be manipulated through the SynCity interface and API. Through this, the client was able to generate synthetic datasets that could test their system and validate its efficacy in edge cases and worst case scenarios.

CVEDIA was also able to provide synthetic segmented LiDAR data within SynCity, minimizing the need for LiDAR data collection, and removing the need for complex, data-intensive segmentation algorithm development. SynCity’s LiDAR modelling is based off of individual commercial sensor models and the output data format is the same for each.

article image
Algorithm Training CVEDIA Algorithm Training Services
Read more
article image
SynCity Use SynCity to develop, train, and validate your computer vision system with custom simulations
Read more
article image
Synthetic Datasets Train your machine learning model on validated synthetic data
Read more
Benefits
  • icon
    Reduce total project expense
    • SynCity reduces or eliminates the need for sensor deployment and field data
    • CVEDIA synthetic data dramatically reduces data collection and storage expenses
    • SynCity retrieves metadata and annotation from synthetic data - an expense you can absolve from your project
  • icon
    Minimize project length
    • Training autonomous applications on synthetic data dramatically reduces time requirements for gathering and organizing field data
    • Datasets can be created and exported in minutes in the browser with SynCity
    • CVEDIA uses assisted algorithm training by choosing which algorithms to apply to most quickly and completely validate your machine learning system
  • icon
    Validate your system effectively
    • Account for unmet field data scenarios that may be too difficult, expensive, or dangerous to explore, and simulate fog, sun glare, dust, sand, and water on your sensors
    • Choose from thousands of 3D models (or work with us to create your own) including infrastructure, natural elements, and humans to ensure your system behaves correctly during unexpected or edge case scenarios

Working with CVEDIA

image
Talk to a team member

CVEDIA projects are developed by an experienced in-house team led by synthetic data industry leaders Arjan Wijnveer and Rodrigo Orph. Our team has been carefully vetted for over 10 years and includes AI veterans with backgrounds in ADAS machine learning, R&D, and large scale deployment.

CVEDIA works iteratively on client projects until satisfactory results are met. Our team has been met with praise from previous and ongoing clients across a range of industries, and we’re happy to be backed by FLIR Systems, the world’s leading thermal sensor producer. CVEDIA works to create custom environments and tool systems for each project, with varying levels of in-house service dependent on client needs.