May 6
You Can Doubt the Safety of Autonomous Driving, but Big Data Tells You: It Does Drive Better than You — Part1
The automobile industry is going through a transitional period from L2 intelligent driving assistance to L3 autonomous driving. The era of intelligent driving is coming. Although every smart driving accident will lead to extensive discussions and concerns about autonomous driving, the numbers will not lie, autonomous driving has greatly reduced the rate of traffic accidents. It is entering a critical point of maturity in technology, testing, laws, and regulations. The automotive powers represented by China, the United States, Europe, Japan, and South Korea are competing to seize the commanding heights of the industry and promote each other to form an existing autonomous driving social governance model with national characteristics and mutual tolerance. The day when humans officially hand over control of vehicles to machines is not far away.
“Autonomous vehicles can effectively avoid 80% of car accidents .”
On December 16, China Automobile Center, Tongji University, and Baidu jointly issued the “Autonomous Driving Vehicle Traffic Safety White Paper,” and the above conclusions were reached through a comparative analysis of authoritative technical demonstrations and actual autonomous driving accidents.
The release of this paper uses data to tell the public a basic fact that although autonomous driving at this stage is not 100% accident-free, autonomous driving is far safer than human driving and has sufficient data support.
On the other hand, with Lidar landing on mass-produced passenger cars, the cameras are becoming clearer, and driving algorithms are iteratively evolving every day. Compared with the individual variations, they are susceptible to various uncertain factors such as emotions and states. The advantages of autonomous driving over humans will become increasingly prominent.
According to the “Statistical Annual Report on Road Traffic Accidents” of the Traffic Administration of the Ministry of Public Security of China, from 2017 to 2019, the country’s traffic accidents occurred an average of 231,900 times each year, with an average annual death toll of 63,000, and another 240,000 non-fatal injuries. Road traffic accidents have become the second leading cause of death among children in the country, and as the only non-disease factor, it ranks among the top ten causes of death in China.
The risk of traffic accidents mainly comes from human drivers.
Traffic accidents caused by human subjective errors accounted for 79.9%.
According to the data of passenger vehicle accidents in the China Automotive Center Institute of Vehicle Safety and Appraisal Technology (CIDAS), from 2011 to 2021,autonomous cars can effectively avoid 80% of accidents caused by human driving.
Therefore, the white paper concludes that the perception function of autonomous driving can detect more than 90% of accidents in advance compared with the limited ability of humans.
Self-driving cars can effectively avoid accidents caused by overspeed, rear-ending, violating traffic rules, and human defects. It can effectively reduce the occurrence of accidents by more than one-third.
In other words, autonomous driving will strictly follow the traffic rules to drive, achieve early perception, and data decision-making process operation.
It is conceivable that in a high-speed scene at night, autonomous driving will detect the vehicle ahead in advance and maintain the distance between cars in strict accordance with safety standards. The probability of an accident will be much lower than that of a human driver.
Therefore, self-driving cars can be aware of other traffic participants to the utmost extent and reasonably maintain a safe distance from other motor vehicles. At this time, the human failure to pay attention to other participants and maintain a safe distance in the cause of an accident can be effectively improved.
Data is meaningful only if it is well labeled
The mainstream algorithm model of autonomous driving is mainly based on supervised deep learning. It is an algorithm model that derives the functional relationship between known variables and dependent variables. A large amount of structured labeled data is required to train and tune the model.
On this basis, if you want to make self-driving cars more “intelligent”, and form a closed loop of the business model for self-driving applications that can be replicated in different vertical landing scenarios, the model needs to be supported by massive and high-quality real road data.
In the field of autonomous driving, data annotation scenes usually include changing lanes and overtaking, passing intersections, unprotected left and right turn without traffic light control, and some complex long-tail scenes such as vehicles running red lights, pedestrians crossing the road, and roadsides as well as illegally parked vehicles, etc.
The current artificial intelligence is also called data intelligence. At this stage of development, the more layers of the neural network, the larger amount of labeled data is needed.
A New Solution For the Self-Driving Data Annotation Project
ByteBridge, a human-powered and ML-powered data training platform provides high-quality services to collect and annotate different types of data such as text, image, audio, and video to accelerate the development of the machine learning industry.
Quality Guarantee
- ML-assisted capacity can help reduce human errors by automatically pre-labeling
- The real-time QA and QC are integrated into the labeling workflow as the consensus mechanism is introduced to ensure accuracy
- Consensus — Assign the same task to several workers, and the correct answer is the one that comes back from the majority output
- All work results are completely screened and inspected by machines and the human workforce
In this way, ByteBridge can affirm our data acceptance and accuracy rate is over 98%.
Communication Cost Saving
On ByteBridge’s SaaS dashboard, developers can start the labeling projects by using the labeling instruction template and get the results back instantly.
From online setting labeling briefing to expert support alongside, the instruction communication is not that hard anymore.
3D Point Cloud Annotation Service
ByteBridge self-developed 3D Point Cloud labeling, quality inspection tool, and pre-labeling functions can complete high-quality and high-precision 3D point cloud annotation for 2D-3D fusion or 3D images provided by different manufacturers and equipment, and provide one-station management service of labeling, QA, and QC.
More info: ByteBridge Launches World’s First Mobile 3D Point Cloud Data Labeling Service
3D Point Cloud Annotation Types:
- Sensor Fusion Cuboids: 49 categories include car, truck, heavy vehicle, two-wheeled vehicle, pedestrian, etc.
- Sensor Fusion Segmentation: obstacles classification, different types of lanes differentiation
- Sensor Fusion Cuboids Tracking
① Tracking the same object with the same ID, labeling the leaving state;
② Time-aligned 2D images could be provided, point clouds outputs only.
Advantages of Our 3D Point Cloud Annotation Service:
· Support 2D/3D sensor fusion, support multiple cameras
· Support scalable data annotation
· AI-powered sensor fusion tool: labeling at 2X-5X speed
· Ease of use QC tool: real-time revision and synchronous feedback
Cost-effective
A collaboration of the human-work force and AI algorithms ensure a 50% lower price compared to the conventional market.
End
If you need data labeling and collection services, please have a look at bytebridge.io, the clear pricing is available.
If you would like to have a look at the 3D point cloud live demo, please feel free to contact us: support@bytebridge.io
Source: https://aikahao.xcar.com.cn/item/1069969.html
