Synthetic Training Data

// Quickly and thoroughly train your artificial intelligence (AI)

rFpro enables the training, testing and validation of a vehicle’s AI through the mass creation of highly accurate training data. It subjects the AI to more edge cases and life-threatening situations than would be possible in the real world, all without compromising safety.

Quickly and thoroughly train your Artificial Intelligence (AI)

rFpro enables the training, testing and validation of a vehicle’s AI through the mass creation of highly accurate training data. It subjects the AI to more edge cases and life-threatening situations than would be possible in the real world, all without compromising safety. This approach more thoroughly exercises the AI more than real-world testing alone. Importantly, human drivers can share the same virtual environment that is being used to train your AI. This injects the unpredictable, random events that provoke new failure modes in autonomous vehicles. rFpro’s complete end-to-end, physically modelled, simulation means it can be used for training AI, regression testing and identifying new failure modes, this enables the simulation to form part of future regulatory frameworks. It can also be used to involve the passengers of autonomous vehicles at the earliest stages of development. This will be critical to ensuring consumer acceptance and protecting your investment in AI and autonomy. Quickly and thoroughly train your AI with rFpro’s Data Farming solution. Creating robust AI requires huge volumes of detailed and diverse training data. rFpro’s synthetic training data production solution – we call Data Farming – helps to rapidly accelerate this process.

Quality training data for autonomous vehicles

rFpro’s Data Farming enables the cost-effective creation of high-quality training data sets that is essential for the development of autonomous vehicles. This simulation-led approach faithfully recreates a limitless number of complex test scenarios using just a single PC, or it can be fully scalable using multiple PCs. 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.

To generate training data for autonomous vehicles the industry has traditionally relied on the manual annotation of real-world video frame, LiDAR points and radar returns to identify objects in a scene. Manually highlighting vehicles, pedestrians, road markings and traffic signals can take 30 minutes per frame and, on average, has a 10% error rate. It takes 1,800 hours to annotate one minute of video. Data Farming generates error-free data up to 10,000 times faster. It creates the required training data quicker, cheaper and more accurately than manual annotation.

This is what rFpro 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”.

Scale training data creation quickly and cost-effectively

Our software enables a range of variables to be simulated, including weather, lighting, traffic and pedestrians. This can all be done using rFpro’s vast digital twin library, which delivers all types of roads and environments from around the world. Each digital twin can produce millions of scene variations, generating enough data to thoroughly train your deep neural network. It simply is not possible to gather the same level of variation in training data in the real world.

Data Farming enables simulation to run faster or slower than real-time depending on the application. Reduced speed running enables complex 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 much faster than real-time.

Read the full article on our Data Farming capability.

ACCELERATE THE GENERATION OF HIGH-QUALITY TRAINING DATA FOR AUTONOMOUS VEHICLES

rFpro’s Data Farming enables the cost-effective creation of high-quality training data sets that is essential for the development of autonomous vehicles. This simulation-led approach faithfully recreates a limitless number of complex test scenarios using just a single PC, or it can be fully scalable using multiple PCs. 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.

To generate training data for autonomous vehicles the industry has traditionally relied on the manual annotation of real-world video frame, LiDAR points and radar returns to identify objects in a scene. Manually highlighting vehicles, pedestrians, road markings and traffic signals can take 30 minutes per frame and, on average, has a 10% error rate. It takes 1,800 hours to annotate one minute of video. Data Farming generates error-free data up to 10,000 times faster. It creates the required training data quicker, cheaper and more accurately than manual annotation.

Contact one of the team to learn more about how rFpro can support your vehicle development projects.