OutStanding DataTech-Intelligent Cabin

Intelligent Cabin Data Service Overview

OSDT possess extensive expertise in delivering AI data annotation services for the multi-intelligent cockpit scenario. Our team excels in collecting and annotating various data types, including speech commands, speaker identification, emotions, sentiment analysis, natural language understanding, audio quality assessment, and command control. With meticulous annotation processes and adherence to strict guidelines, we ensure the provision of accurate and reliable labeled datasets for training AI models. Our proficiency in this field enables us to provide high-quality annotations, contributing to the advancement of voice recognition, cockpit control, and improved user experiences within the multi-intelligent cockpit environment.

gesture interaction, the adoption of three-dimensional dynamic gestures and complex gestures is also accelerating its application.

Case study: NLU annotation Implementation Key Points

Project overview

The basic idea of annotation is to classify the utterance into categories. And utterances are typically things that people have said to control their smart in-cabin system. The annotation process is done in three phases: Domain classification Parsetype labelling Mention labelling

Annotation example

find the movie [avatar]

Domain: visual_media

Parsetype(including intent and object): the intent is search and the object is movie

Mention: [avatar]<movie>

Intelligent Cabin

Case study: OMS annotation Implementation Key Points

Project overview

In the images extracted from videos, people should be selected and annotated, and 18 key points of all the passengers in the car should be labeled. For partially displayed areas, their positions are inferred using imagination.

Case stuy: In-cabin commands generalization Implementation Key Points

Project overview

The client provides the original control commands. Based on the language usage habits of the local country, resources should understand the intent of the initial text and generate the synonymous extension. Quality is the top priority instead of quantity for each intent.