Started the sale for public road data set for deep learning research and development
- Provide public road image data with annotations in multiple methods -
An example of annotation data (※ upper and lower original images are different)
Upper figure: bounding box method with object enclosed in rectangle,
Lower figure: Semantec segmentation method in which objects are painted in pixel units
ZMP Inc. (Bunkyo-ku, Tokyo, Representative Director: President: Hisashi Taniguchi, hereinafter referred to as ZMP) began the sale of deep learning set for research and development which is available as a menu of service data acquisition solution RoboTest®.
As the revitalization of the ADAS market and the development of Autonomous Driving technology are progressing, attempts to increase the recognition accuracy of travel-able areas, preceding vehicles, surrounding vehicles, pedestrians, signals, signs, white lines, etc. utilizing deep learning are prosperous. In image deep learning, it is necessary to accurately label objects in the image and road areas as learning data.
The annotation data which is to be sold this time is carrying out various labeling as the data set for learning on the data we acquired by public road driving. Traveling scenes include urban areas and highways, not only straight lines but also data including lane changes and intersection images. Classification of labels is classified as automobiles, motorcycles, people, bicycles, signs, traffic lights, etc. Utilizing these data enables customers to check data at the initial stage of R & D such as deep learning. Also, you can easily obtain and use it as a sample for creating your own annotation rule.
Thanks to the provision of this data, we will support customers' acceleration of research and development by acting on behalf of customers to prepare equipment necessary for data creation, data acquisition by running, annotation processing and so on.
The driving data acquisition solution RoboTest® not only acquires the customer's original driving image data, but also acquires driving data along with vehicle information. We also offer a data set creation service based on customized annotation rules.
In addition, we are planning to provide a tool that can automatically perform online viewer and scene classification that simultaneously displays the acquired data as a robot data platform.
【Create and sell public road data set for deep learning research and development/sales service】
【About RoboTest® for acquisition of driving data】
Since 2012, ZMP is using data from our own Autonomous Driving technology development and testing to provide data acquisition support services during driving by public drivers. In addition to collecting and managing drivers necessary for traveling, upstream processes such as the design and construction of systems (loggers) necessary for acquiring travel data, vehicle arrangements and necessary installation / remodeling, and data processing after acquisition,it is a service that can drastically reduce customer's labor by implementing it all at once.
<Contents of RoboTest® service>
① Consulting for acquiring travel data
② Development of data acquisition system (hardware, software)
3. Operation management of recruitment and management of drivers, data acquisition
④ Data postprocessing (tagging, format conversion etc. according to customer requirement specifications)
⑤ Analysis and report of acquired data
For details of RoboTest®, please visit the following website.
【Data Analysis / Analysis Platform Service Robo Data Platform】
A new tool were data analysis users can freely apply data learning such as deep learning to machine learning, object recognition processing, tagging by scene from sensor information.We provide a data post-processing service environment that can analyze various things such as emotional analysis of the driver. This service reduces the time it takes to search for traveling data from tagged information and streamlines analysis work.
For details of Robo Data Platform,please visit the following website.https://www.zmp.co.jp/products/robotest/rdp
Please contact us for price.
【Press release article】