Platform

Overview

Developed by ECI with cutting-edge models, this platform brings together native speakers of 300+ languages who are available around the clock to provide you with top-quality training data for machine learning.

It consists of three components:

• Web Data Management Console
• Web Data Annotation Center
• WeChat Mini Program for Data Collection (WiiWork)

Data Annotation Capabilities

Annotation Management

The Data Annotation Center allows customized configuration of data types and annotation attributes to cater to a diversity of client needs. Its intuitive interface enables project managers to keep track of task status and progress in real time and settle up with task teams in a timely manner.

Production of Annotated Data

With the Web Data Annotation Center, we produce annotated data by labeling the data available in various formats like text, video, audio, or images.

Data Collection Capabilities

Resources Management

Our data collection is based on the substantial resources accumulated by ECI’s crowdsourcing platform, which form a resource pool after our internal identification and evaluation.

Data Collection

Powered by the mobile Internet, our WeChat Mini Program WiiWork makes it easy for us to collect data in various formats like image, audio and text through qualified crowdsourcing resources anywhere and anytime.

Point Cloud Annotation

Lidar-Cameras are indispensable sensors for autonomous driving. 3D point clouds and 2D images files are important data collected by these sensors. Advanced 3D point cloud processing and multi-sensor fusion technology are cornerstones for autonomous driving. With cutting-edge point cloud algorithms, our platform can help car companies explore the boundary of LIDAR, bringing new vitality to autonomous driving.

The platform supports annotation of sensor files with different devices and different data densities. Thanks to 3D vision technology, we can provide solutions in

General point cloud annotation:

1. Point cloud noise reduction, ground pre-processing, minimization of target fitting and automatic target orientation prediction with AI
2. Annotation fine-tuning, annotation semanticization, single-target focus, background off, etc.
3. Bidirectional real-time fusion-based annotation of multi-sensor data

Continuous frame annotation: Based on cutting-edge target tracking algorithms, the size and translation of the target can be estimated in both offline and online scenarios, thus ensuring the consistency of the trajectory and achieving continuous frame annotation of the target. This method can save the annotation time by more than 85%, thus significantly increasing the degree of automation.

Cross-frame copy: mainly used for copy and paste of target size types in cross-frame situations, assisting target replenishment and marker matching