The Embodied Eye: Aurora S Unveiled

Hello everyone!
We often say that for machines to truly understand the world, they first need a good pair of “eyes.” Today, Xiaolan takes you on-site to witness how Aurora S, beyond providing powerful spatial localization and mapping capabilities, also becomes the “visual powerhouse” of robots through sheer technological strength!

Highlight 1: 120° Ultra-Wide Stereo Depth Perception — Far Beyond Traditional Solutions

As the core sensor for robot obstacle avoidance, the depth camera’s field of view (FOV) directly determines how wide a robot can “see” during motion. Unfortunately, traditional depth cameras often have limited FOVs, creating blind spots in obstacle detection.

In Aurora S, we’ve innovatively implemented an end-to-end neural network that achieves a 120° ultra-wide-angle depth perception — wide enough to cover the entire area in front of the robot.

We placed Aurora S and a conventional depth camera in the same position to capture the same scene. Let’s see how they compare.

Third-person perspective shooting

The result? A striking difference! Thanks to its 120° horizontal FOV, Aurora S captures a much broader area in a single shot, taking in the full environmental picture. In contrast, the traditional camera’s view looks noticeably cramped.

Traditional camera vs. Aurora S depth comparison

Indoor shot of same object — Aurora’s FOV is wider

In short: Robots and intelligent devices can now gather more comprehensive environmental data at a glance — enabling faster decisions and safer actions.

Highlight 2: The “Weak Texture Terminator” — No Fear of Black Counters or Elevator Doors

Many depth cameras fail when facing smooth, low-texture surfaces. So, we ran a series of tough weak-texture tests:

1. Black Counter Test: The traditional camera’s depth map shows obvious “holes” and missing data, while Aurora S delivers a smooth, complete, and stable output.

Traditional vs. Aurora S depth comparison

Counter overlay mode

2. Elevator Door Test (Stainless Steel): Traditional cameras almost fail entirely on reflective metallic surfaces. Yet Aurora S clearly reveals the depth contrast and 3D structure of the elevator door.

Traditional vs. Aurora S depth comparison

Elevator recognition overlay mode

The result: Whether in a dim café or a shiny elevator lobby, Aurora S provides reliable 3D perception — effectively eliminating visual blind spots.

Highlight 3: Excellent Performance Under Bright Outdoor Light

Bright outdoor light and long-range distances are traditionally nightmares for depth cameras. But Aurora S, powered by its end-to-end deep learning architecture, shines in these conditions. Here’s what we found:

1.Dynamic Capture on the Move

During outdoor walking tests, the Aurora S depth map remained stable and continuous — no tearing or data loss during rapid movement. The 3D structure of the environment was reconstructed in real time.

Traditional vs. Aurora S depth comparison

Outdoor walk overlay mode

2. Accurate Roadside Detail Detection — Outdoor Safety Enhanced

We tested metal guardrails along the road. Traditional depth cameras produced sparse or missing data, but Aurora S sharply outlined the continuous structure of the rails, providing vital perception for stable outdoor obstacle avoidance.

Traditional vs. Aurora S depth comparison

Guardrail overlay mode, long view

In essence: Aurora S enables robots to operate seamlessly from indoors to outdoors, maintaining clear 3D perception even under strong light or dynamic movement.

Highlight 4: Dynamic Semantic Recognition — Seeing Not Just “Where,” But “What”

Aurora S doesn’t just measure distances — it helps machines understand their surroundings.

Outdoors: It distinguishes over 18 types of scene elements, including buildings, roads, vegetation, and vehicles.

Indoors: It accurately recognizes over 80 types of common objects, such as people, tables, chairs, and monitors.

Its recognition capability is incredibly fine-grained — even detecting a person from just an arm or leg, and identifying the “phone” in someone’s hand. This leads to strong occlusion resistance and opens up rich interaction possibilities.

Highlight 5: 3D Semantic Point Cloud — From 2D Understanding to Full Spatial Awareness

Aurora S projects all recognition results directly and accurately into a 3D point cloud in real time.

In this 3D world, you no longer see random points, but clearly color-labeled “roads,” “vehicles,” “pedestrians,” and “buildings.” This empowers robots to make decisions and interact directly in 3D space.

Seeing Is Believing: Real-Time Office Demo

We conducted a real-time demo of Aurora S in an office setting.

As shown above, when connected via Remote UI, you can observe two live modules on the main interface:

Depth Perception:Accurate 3D geometry of the environment

Semantic Recognition: Precisely color-labeled object categories

All in real time — demonstrating Aurora S as a powerful and ready-to-use 3D scene understanding solution.

In Summary

Aurora S is not just a depth camera — it’s a comprehensive visual hub for intelligent machines.

With its:

• 120° wide-angle neural depth perception— seeing more,
• Anti-texture weakness— seeing stably,
• Strong light resistance— seeing farther,
• AI semantic segmentation & 3D semantic projection— seeing and understanding better.

We believe powerful tools spark endless innovation.

May Aurora S become the most reliable “pair of eyes” behind your next breakthrough.