Success Story Manufacturing

Automotive OEM Gets Accurate Video Labelling for Autonomous Vehicles

Download
Faster mobility application development for Automotive

With autonomous car technology cresting the horizon, automakers are working with IT partners to drive greater innovation and safety margins within every self-driving vehicle.

One of Germany’s most well-known OEM’s launched a new technology division aimed at perfecting autonomous driving technology. The firm looked to Apexon to help them with a video labelling solution that would tag various objects in and around a car, enabling faster mobility application development.

the Results

Key Outcomes

Apexon successfully hit every milestone and reached all mandated targets over the course of two years, processing over 25 million labels in that time frame. Additionally, we were able to:

Consistently deliver an SLA target of 99.5%, with under 5 defects per 1000 labels
Consistently deliver an SLA target of 99.5%, with under 5 defects per 1000 labels

Drastically reduce project hiring times & infrastructure set up costs
Drastically reduce project hiring times & infrastructure set up costs

Deploy a reliable data security framework with scalable options for future business growth
Deploy a reliable data security framework with scalable options for future business growth

Establish stringent resource management protocols, including a 10% resource buffer
Establish stringent resource management protocols, including a 10% resource buffer

The challenge

5 key areas

While video labelling is not an entirely new technology, this particular project posed certain significant challenges.

Increasing Costs

Quality Solution Delivering a quality solution with a very low scope for error and zero timeline to rectify errors

Poor Engagement

New Work Types Handling new work types with relatively unknown business rules and volume flow

Management Processes

Management Processes Low resource availability, high attrition rate, and short annual hiring timelines

Building an ODC setup

ODC Setup Building an ODC setup to work with patented technologies and solutions

Poor Video Files

Poor Video Files Working with poor quality video files

The solution
4 key areas

In building and deploying the solution, Apexon engineers used several tools to annotate the video labels and streamline the hiring process. Solution details include:

Using Appropriate Toolsets & Programs

Using Appropriate Toolsets & Programs Using a blend of proprietary toolsets and Philosys Label Editor to fully annotate objects using textural and geometric markers and classes, while delivering them in XML data format – to be used in algorithmic validation and machine learning applications

Cvedia Platform

Cvedia Platform Leveraging the Cvedia platform to enable deep learning vehicle applications via live streaming video footage, and Vatic annotation tool to quickly build massive video sets that are deployable to Amazon Cloud

Progression Indicators

Progression Indicators Deploying outcome-based incentive models and cross training as a career progression indicator

New Hiring Processes

New Hiring Processes Tailoring new hiring processes to meet the requirements specific to the received work input format