Friday 07 March 2025
As robots become increasingly sophisticated, their ability to perceive and interact with their environment is crucial for successful task completion. One of the most important sensors in a robot’s arsenal is the depth sensor, which provides information about the distance between objects and the robot itself. However, choosing the right depth sensor can be a daunting task, especially when considering the many options available on the market.
Researchers at the Czech Technical University in Prague set out to address this issue by comparing four state-of-the-art stereoscopic depth sensing cameras: Intel RealSense D435, Intel RealSense D455, StereoLabs ZED 2, and Luxonis OAK- D Pro. These cameras use a similar principle to work, but differ in their design, capabilities, and performance.
The researchers tested the cameras using three different scenarios: perceiving planar surfaces, a plastic doll, and objects from the YCB dataset. They recorded over 3,000 frames for each camera and evaluated its performance using six metrics: bias, standard deviation, chamfer distance, Jaccard similarity, F1 score, and angle between normals.
The results show that even though all cameras have similar principles of operation, they perform differently in different contexts and distances. The Intel RealSense D435 performed well at lower distances, but struggled with more complex objects. The Intel RealSense D455 worked better on planar surfaces and larger distances, but had a lower RGB resolution than the other cameras.
The StereoLabs ZED 2 camera provided the best overall performance, working well in all three scenarios and providing easily accessible AI features like keypoint detection. However, it requires a CUDA-enabled GPU to function, which may be a limitation for some applications.
The Luxonis OAK- D Pro camera performed well on planar surfaces, but struggled with more complex objects. It also has the ability to compute depth estimation onboard without requiring an external GPU, making it a good option for certain applications.
Overall, this study provides valuable insights into the performance of different stereoscopic depth sensing cameras and helps researchers and developers make informed decisions when selecting a camera for their application. The results highlight the importance of considering the specific requirements and constraints of the task at hand, as well as the limitations and strengths of each camera.
For example, if you need to perceive planar surfaces or work with larger distances, the Intel RealSense D455 may be a good choice.
Cite this article: “Comparing State-of-the-Art Stereoscopic Depth Sensing Cameras”, The Science Archive, 2025.
Robotics, Depth Sensors, Stereoscopic Vision, Camera Comparison, Intel Realsense, Stereolabs Zed 2, Luxonis Oak-D Pro, Czech Technical University, Prague, Ycb Dataset, Computer Vision.
