On April 16, JD.com officially launched the JoyEgoCam ultra-high-definition data collection terminal at the Embodied Intelligence Ecosystem Launch Event. As a hardware product specifically developed for AI embodied intelligence training, JoyEgoCam not only marks the application of ultra-high-definition data collection technology, but also becomes the world's first embodied intelligence data infrastructure that covers the entire "data collection, storage, labeling, training, evaluation, simulation, and testing" workflow, aiming to establish a complete closed-loop path from physical world data acquisition to model simulation and evaluation.
In terms of technical specifications, JoyEgoCam demonstrates strong industrial-grade performance. The device is equipped with a dual-camera system, achieving an ultra-wide FOV of 130 degrees both vertically and horizontally, enabling the recording of high-resolution video at 2160x2160 with 60 FPS. This combination of high frame rate and high resolution ensures that machines can capture extremely subtle motion trajectories and environmental changes during the learning process, meeting the strict requirements for complex data collection and high-precision simulation and evaluation in embodied intelligence.
Strategically, JD.com has already launched the largest data collection initiative in human history last month. The plan has clear goals: to increase the scale of real-world video data to over 10 million hours within two years. In the context of the current AI competition shifting from parameter scale to high-quality data quality, JD.com's move aims to build a solid technological moat for the widespread commercialization of embodied intelligence by following the logic of "hardware defining data and data feeding models." This not only enhances the generalization ability of underlying algorithms but will also further accelerate the application of robot technology in warehousing, logistics, and multi-scenario services.
