Data Farming | rFpro

Quicker, more cost-effective delivery of high-quality training data for the autonomous vehicle industry.

A significant challenge faces the autonomous vehicle industry regarding the design and implementation of new technologies. Generating training data for deep learning from real-world recorded scenarios is expensive, time inefficient and inflexible. rFpro Data Farming provides an attainable solution that can streamline the vehicle development processes in a shorter time frame and at competitive cost.

Hear what our customers are already saying about Data Farming: 

DENSO ADAS Engineering Services, global Tier1 supplier

“Through rFpro’s Data Farming we can create an extensive number of driving scenarios, allowing the generation of very large variations in scenes, all through the investment in a single platform,” said Francisco Eslava-Medina, Project Manager at DENSO ADAS. “This allows us to quickly and cost-effectively generate the vast quantity of quality training data that is essential for certain product development phases of computer vision technologies, especially for neural networks for our autonomous vehicle technologies.”

Ambarella, leading autonomous vehicle technology provider

“The software presents a radical shift in creating training data and is already accelerating the development of our autonomous vehicle systems,” said Alberto Broggi, General Manager of Ambarella’s division in Italy. “Deep learning and AI are critical to the successful adoption of autonomous vehicles. It may not be reasonably possible to get to the standard required only through the use of manually annotated data sets. Data Farming will transform the way the industry develops autonomous vehicles”.

Watch our video to learn more about this revolutionary approach:

Video also available in German and Japanese.  

 

Reduced hardware costs

The autonomous vehicle industry currently relies on manual annotation of test data – the capture of which is an expensive and time consuming process – created frame by frame, which is susceptible to compromise due to human error. The traditional method of data capture requires a team of specialists manually annotating video frames, LiDAR points or radar returns to identify objects required for training data generation. For example, other vehicles, pedestrians, road markings and traffic signals. This can take 30 minutes per frame with a 10% error rate.
Our new method generates error-free data up to 10,000 times faster compared to manual annotation. This significantly reduces the time pressures, risk of error and man hours required to provide manufacturers and system providers with usable data.

 

Data synchronisation for the most complex hardware designs

We compare our new Data Farming software to the Render Farming technique, which revolutionised the way modern popular animation is compiled – making the creation both faster and more cost effective.
Our software enables the simulation of all test condition variables, from weather to lighting, traffic and pedestrian numbers and behaviour. Such rigour would not be possible with physical testing alone, which itself requires real vehicles operating in limited conditions over a long period of time.

Additionally, our vast digital twin library delivers the world to the development teams, with a variety of off-the-shelf models, including city streets, highways, rural and mountain roads. Each digital twin can produce millions of scene variations, mitigating against over-training the customer’s deep neural network. For more information on our digital twins please click here.

Our customers can compile complete data sets across an entire vehicle with every sensor simulated simultaneously. Data can be fully synchronised even for the most complex hardware designs. This is crucial where users are compiling data from several 8K HDR cameras, LiDAR and radar sensors.

This simulation-led approach enables the recreation of complex test scenarios using a single PC, making high-quality training data attainable where it was previously time- or cost-prohibitive. Data Farming is fully scalable, meaning you can accelerate data production to suit your requirements without considerable additional investment.

 

Streamlined development process

Data Farming helps achieve more accurate results, faster, for the benefit of accelerated vehicle development. Our new method enables simulation to run faster or slower than real time depending on application: for example, where there is no requirement for Driver or Hardware in the Loop (DIL/HIL).

Reduced speed running enables tests to be conducted using fewer processors – or processors of lower specification – to produce repeatable, high quality data. Simple, single sensor data, for example, can be run at much faster than real time.

Data Farming is an entirely tailorable process that can be altered to meet customers’ specific needs. If you would like more information, or to speak to one of our specialists, about how Data Farming can help you, please click here to get in touch.

 

Not what you were looking for?