18080482. METHOD AND ELECTRONIC DEVICE FOR 3D OBJECT DETECTION USING NEURAL NETWORKS simplified abstract (Samsung Electronics Co., Ltd.)

METHOD AND ELECTRONIC DEVICE FOR 3D OBJECT DETECTION USING NEURAL NETWORKS

Organization Name

Samsung Electronics Co., Ltd.

Inventor(s)

Danila Dmitrievich Rukhovich of Moscow (RU)

Anna Borisovna Vorontsova of Moscow (RU)

Anton Sergeevich Konushin of Moscow (RU)

METHOD AND ELECTRONIC DEVICE FOR 3D OBJECT DETECTION USING NEURAL NETWORKS - A simplified explanation of the abstract

This abstract first appeared for US patent application 18080482 titled 'METHOD AND ELECTRONIC DEVICE FOR 3D OBJECT DETECTION USING NEURAL NETWORKS

Simplified Explanation

The abstract describes a method for 3D object detection using a neural network. Here is a simplified explanation of the abstract:

• The method involves receiving one or more monocular images.
• The monocular images are passed through a 2D feature extracting part of the neural network to extract 2D feature maps.
• An averaged 3D voxel volume is generated based on the 2D feature maps.
• The averaged 3D voxel volume is passed through an encoder of a 3D feature extracting part of the neural network to extract a 2D representation of 3D feature maps.
• The 2D object detection in a Bird's Eye View (BEV) plane is performed by passing the 2D representation of 3D feature maps through the outdoor object detecting part of the neural network.

Potential applications of this technology:

• Autonomous driving: This method can be used in self-driving cars to detect and track objects in a 3D space, enabling safer navigation.
• Robotics: It can be applied in robotics for object detection and recognition in a 3D environment, enhancing the capabilities of robots.
• Surveillance systems: The method can be used in surveillance systems to detect and track objects in a 3D space, improving security and monitoring.

Problems solved by this technology:

• Accurate 3D object detection: The method provides a way to accurately detect and track objects in a 3D space using monocular images.
• Efficient feature extraction: The use of 2D and 3D feature extracting parts in the neural network allows for efficient extraction of relevant features from the input images.
• Bird's Eye View object detection: The method performs object detection in a Bird's Eye View (BEV) plane, which can provide a more comprehensive view of the objects in the scene.

Benefits of this technology:

• Improved safety: Accurate 3D object detection can enhance the safety of autonomous vehicles and robotics systems by enabling them to detect and avoid obstacles.
• Enhanced situational awareness: The method provides a detailed understanding of the 3D environment, allowing for better decision-making in various applications.
• Real-time processing: The efficient feature extraction and object detection process enable real-time processing, making it suitable for time-sensitive applications.

Original Abstract Submitted

A method of 3D object detection using an object detection neural network includes: receiving one or more monocular images; extracting 2D feature maps from each one of the one or more monocular images by passing the one or more monocular images through a 2D feature extracting part of the object detection neural network, generating an averaged 3D voxel volume based on the 2D feature maps, extracting a 2D representation of 3D feature maps from the averaged 3D voxel volume by passing the averaged 3D voxel volume through an encoder of a 3D feature extracting part of the object detection neural network, and performing 3D object detection as 2D object detection in a Bird's Eye View (BEV) plane, the 2D object detection in the BEV plane being performed by passing the 2D representation of 3D feature maps through thane outdoor object detecting part of the object detection neural network.