On May 31, 2021, it has been demonstrated that Didi Auto Driving completed a new round of strategic financing for more than $ 300 million, and the investor was the GAC Group.
Didi has raised a total of more than $ 1.1 billion since its demerger in 2019.
In 2016, Didi began developing and testing automatic driving.
In August 2019, Didi upgraded its automatic driving department to a company. On May 29, 2020, Didi travel announced that its independent driving company had completed the first round of financing of more than $ 500 million, led by Phase 2 of the Softbank Vision Fund.
On January 28, 2021, the Didi autopilot completed $ 300 million in funding. This round will be led by IDG Capital, followed by CPE, Paulson, Russian Chinese Investment Fund, Guotai Junan International, CCB International and other investment institutions.
It is said that after the new round of funding, the evaluation of the Didi automatic run will exceed the Pony AI. Artificial intelligence is one of the most powerful automotive vehicles in the industry in China. It has been shown that as of February 2021, Pony AI has raised a total of more than $ 1.1 billion and is worth more than $ 5.3 billion.
In June 2020, the Didi automatic drive announced it would conduct a manned test in Shanghai. It currently has driving licenses in Shanghai, Beijing, Suzhou, Hefei, California, etc. According to Didi, its autopilot R&D department currently has more than 500 people covering road condition detection, accuracy mapping, behavior prediction, planning and management, infrastructure and simulation, cloud control and vehicle Internet, autopilot timing products and user groups.
Mr.Cheng, CEO of Didi Travel, said that by 2025, the Didi platform is expected to favor more than a million units of shared cars, and an iterative version of the new models will be able to carry the driver module developed by Didi himself. By 2030, Didin’s customized cars hope to be completely independent.
According to an official report, there are nearly 5,200 autopilot-related companies in China. Of these, more than 30% of the registered capital is more than 10 million, and almost 90% of them are joint-stock companies.
From the point of view of industrial distribution, companies involved in automatic driving are mostly distributed in wholesale and retail trade. More than 1,500 companies, accounting for 31%, are followed by scientific research and technical services, data transmission, software and IT services, 29% and 20% respectively.
In terms of geographical distribution, Guangdong has the most independent time-related companies, more than 1,600 companies, 32%. In addition, there are more than 300 independent time-related companies in Hebei, Jiangsu and Shandong Province.
Overall, over the past decade, the annual number of registrations of companies related to autonomous cars increased overall, with an annual growth rate of over 17%, and the total number of registrations grew rapidly. Of these, nearly 1,200 new related companies were added to the list in China in 2019, the year with the highest annual registration volume. In 2020, there were nearly 900 new related companies in China. On May 27, 2021, China has added more than 300 related companies this year.
Self-propelled technology is changing the transportation industry, social life and everyday life. It’s hard to know when that day will arrive. Because life is priceless, we must seek perfection from the beginning.
It is challenging for self-driving manufacturers to respond internally to the growing demand for high-quality data labels.
A tool assisted by artificial intelligence
The stamp used markings point by point, which cost a lot of time.
3D marking and video recording considered to be the most difficult services for data entry. Currently, object-tracking algorithms based on machine learning have already helped with video annotations. The annotator marks the objects in the first frame and then the algorithm tracks those in the following frames. The annotator only needs to change the entry when the algorithm is not working well. It’s 100 times faster than before.
Thanks to the artificial intelligence-assisted system, the corresponding parts can be automatic and can be transcribed, and one only needs to check and edit the wrong parts.
Today, some artificial intelligence-assisted tools come to the fore in practice in two factors.
- Reduce costs: Artificial intelligence-enabled features allow customers to save more money as labor costs fall.
- Shortening time: Make a large-scale requirement for training data in a short time. Using an artificial intelligence tool can improve efficiency many times over
Can we get rid of the human workforce?
The answer is no.
In fact, manually tagged data is less prone to errors in quality assurance and data exceptions.
The human workforce cannot be completely replaced by some tools that work with artificial intelligence-based automation features, especially with exceptions, edge cases, complex data tagging scenarios, and so on.
As mentioned, data accuracy is vital in the automotive industry, here comes a new question: Should I build an internal team?
Before making a final decision, you need to keep 2 things in mind:
1 Complex process: including labeling tools and built-in data for data pre-processing, training and monitoring of label performance, data validation and quality control, etc.
2 High financial participation: such as infrastructure labor costs, research and development, etc.
Compared to the in-house infrastructure outsourcing service needs effective communication and quick feedback. It is very important for manufacturers to choose the right one that can act as an “expansion department” for their business.
The following elements must also be taken into account:
Progress preview: customers can monitor the progress of entries in real time on the dashboard
Result preview: customers can get results in real time on the dashboard
Customer service: customers can communicate with data employees about changes so employees can react quickly and make changes to the workflow
Finally, in the Autocar industry, we rely a lot on people’s workforce. Therefore, when choosing an outsourcing partner, we need to ensure flexible engagement in the labeling loop because we need labels to respond quickly and make changes to the workflow based on the testing and validation phase of the model.
Quality and accuracy
- ML-assisted capacity can help reduce human error automatically advance markings
- real-time quality control and quality control integrated into the labeling workflow when a consensus mechanism is put in place to ensure accuracy.
- Understanding – Assign the same task to multiple employees, and the correct answer is the one that comes back from the majority output.
- All the results of the work has been fully security checked and inspected machine and manpower.
In this way, ByteBridge can confirm the acceptance and accuracy of the data over 98%
- The customer can directly set the labeling rules is dashboard.
- The customer can replicate data properties, attributes, and task flow, expand or collapse, make changes based on what they learn about model performance at each stage of testing and validation
For example, you can select Autopilot marking model for your project:
- Preview progress: the customer can monitor the progress of the entries in real time on the dashboard
- Preview the result: the customer can get the results in real time on the dashboard
Real-time outputs: the customer can get through real-time outputs API. We support JSON, XML, CSV, etc., and we can provide editable data type to meet your needs.
These marking tools are already available on the dashboard: Classification of images, 2D boxing, Polygon, Cuboid.
3D Point Cloud Markup Service
ByteBridge has self-developed 3D Point Cloud markings, a quality control tool, and label functions that can complement high-quality and accurate 3D point cloud markings for 2D-3D fusion or 3D images provided by a variety of manufacturers and devices, and provide a single station management service and quality control
3D point cloud annotation types:
- Sensor Fusion Cuboids: 12 categories include car, truck, heavy vehicle, two-wheeled vehicle, pedestrian, etc.
- Sensor fusion segmentation: barrier classification, different types of band specifications
- Monitoring of sensor fusion cuboids
① Tracking the same object with the same ID, marking the initial state;
② Dot clouds or time-aligned images can be output, only dot cloud outputs.
Cooperation between man and the workforce Artificial intelligence algorithms ensure a 50% lower price compared to the normal market.
Because the quality of the marked data sets determines the success of the self-driving industry, collaboration reliable partner can help developers overcome data labeling challenges.
We can offer personal marking tools and services according to customer requirements.
If you need data entry and collection services, check out bytebridge.io, clear pricing is available.
Feel free to contact us: email@example.com