Seoul-based INFINIQ, which specialises in artificial intelligence and machine learning data services, is expanding fast. It’s just revealed that it will be opening a dedicated European headquarters to expand the scope of its business beyond South Korea to the global autonomous driving data market.
The company provides AI data sets, such as vision, voice, and text, to companies that develop deep learning applications such as autonomous driving, face recognition, among a host of other solutions.
Keeping in line with its agenda of expanding its presence to other geographies, INFINIQ was also a part of CES 2021, where it lifted the covers off a brand-new cloud sourcing platform called MYCROWD, as well as data service for autonomous driving.
To find out more about these developments, Auto Futures spoke to INFINIQ’s Chief Marketing Officer, Gyeseung Kim.
“INFINIQ is a long-established company with nearly 17 years of history and has full domain knowledge of ADAS, telematics and autonomous driving with Software Validation,” says Kim.
“Initially we were doing software validation which is service to diagnose and improve problems of SW and HW operating products and services but have since transitioned into providing datasets for AI training, where we take advantage of our expertise in quality management.
“In terms of training data for AI, it is a process of converting unstructured data into structured data for AI model development. For that, our service is specialised in data collection and annotation for AI model training, and in particular we focus on data for autonomous driving fields, which need more high quality than others,” he adds.
INFINIQ operates specialised vehicles to collect data for autonomous driving training.
“This vehicle has high resolution vision cameras (covering 360° views), LiDAR which collects 3D cloud data with location information, infrared thermal and night vision cameras to cover night shooting and prepare for bad weather. With these features, environmental control and storage systems are installed in the vehicle, as well to respond and meet requirements from various clients, who develop autonomous driving systems and algorithms.”
After collecting raw data, it has to process for data curation, which can be done automatically with its tool which selects optimal data for the purpose of training.
“What we make different from other competitors is 3D LiDAR sensor fusion technology. There is a lack of global professional manpower as well as skilled experts in 3D data processing. Some companies are promoting 3D data sensor fusion processing, but there are many cases that cannot be confirmed. However, INFINIQ has accurate calibration know-how between 2D and 3D LiDAR data, as well in experience in 3D Point Cloud data: object recognition, auto-labelling and sensor fusion,” explains Kim.
Speaking about the data annotation side of things, he answers: “INFINIQ has a self-developed annotation tool to flexibly respond to various requests and specifications from customers. It can be magnified 50 times and magnified by 1 pixel for high point accuracy. Setting curves rate for each project is acceptable in our tool. Furthermore, drawing types except normal function in the annotation tool and classification per domains are available to respond to variations in annotation. This degree can also be defined for each picture type or domain that the annotation tool has.”
MYCROWD is a global crowd workplace for AI data annotation assisted by awesome ‘magic’ functions.
INFINIQ’s impressive range of solutions focus on three key areas – auto object detection, de-identification and auto-labelling. For starters, its auto object detection automatically detects objects and accurately draws an outline. It is detectable for all objects based on its customers’ needs.
Its de-identification solution recognises accurate areas, does not affect AI learning effect, meets GDPR requirements through data compliance and protection, and recognises all races, wearing masks, number plates from any countries.
And finally, its auto-labelling solution efficiently recognises and labels more than 300 object types. It also intelligently selects and annotates specific objects based on what their customer wants and can be combined with the annotation tool to improve productivity.
Speaking about INFINIQ’s other offerings, Kim told us more about the MYCROWD platform.“MYCROWD is a global crowd workplace for AI data annotation assisted by awesome ‘magic’ functions.”
“It utilises deep learning technology and has the advantage of quickly recognising all objects by using unique auto object detection function and being able to use it for multiple purposes such as annotation. Auto object detection has automatic pre-processing functions such as auto rectangle (Magic Fit) and auto polygon (Magic Pin). MYCROWD realizes more productivity and easier work for annotators. Customers can obtain large amount of data with high quality for AI training.”
Adding on about the company’s work in the HD Mapping space, he says: “In the case of indoor and outdoor HD Mapping, in addition to collecting data (eg: LiDAR, GPS), post-processing processes such as road surface processing, data clustering, shape estimation (layer classification process) and noise removal are involved. To summarise, we can provide all services in the areas developing data acquisition equipment and setup, data collection and post-processing for HD Mapping.”
INFINIQ’s AI driven data sets help develop some aspects of the automotive ecosystem like AVN, Cluster, HUD, Infotainment, Telematics, HVAC systems. Speaking about them, Kim says: “When it comes to the ADAS system, INFINIQ’s AI driven data sets are helping the learning and advancement of AI algorithms for driving, parking, road surface recognition system as GT Data.”
With INFINIQ expanding its presence in Europe, I wanted to know what role this new establishment would play to expand its contribution to autonomous driving, and what the future looks like for the company.
Kim replies: “Luxembourg, where our new office is located, is the place to support research and innovation activities for automobile-related companies and is a strategic point to letting traveling within an hour to all regions of Europe countries such as Germany, France, Sweden and Italy. Based on this, we will accelerate its full-fledged global advancement in training data for autonomous driving field. We plan to increase corporate investment efficiency and strengthen R&D cooperation through information sharing with active networking.”
“As for the future, we plan to expand our business based on domestic car manufacturers and autonomous driving companies, and based on this, we plan to secure global car manufacturers, thereby preoccupying our position as a leading data provider for autonomous driving learning. We aim to increase the proportion of overseas sales to more than 50% of total sales by 2022.”
“We also plan to commercially launch a movable concept to acquire driving data for the global market. Moreover, the crowd platform MYCROWD is expected to expand to begin services globally,” he concludes.