Methods for performing object recognition and / or semantic segmentation for a vehicle
DE102025104147A1Undetermined Publication Date: 2026-08-06VALEO SCHALTER & SENSOREN GMBH
Patent Information
- Authority / Receiving Office
- DE · DE
- Patent Type
- Applications
- Current Assignee / Owner
- VALEO SCHALTER & SENSOREN GMBH
- Filing Date
- 2025-02-05
- Publication Date
- 2026-08-06
Smart Images
Figure 00000000_0000_ABST
Abstract
The invention relates to a method for performing object recognition and / or semantic segmentation for a vehicle (1), comprising providing (S1) a color image (10) and an infrared image (11); and applying (S2) a neural network module (12) to the color image (10) and the infrared image (11). The neural network module (12) comprises a backbone network (13) that determines multiple feature maps (14) with different resolutions; a multi-scale feature fusion network (15) that receives the multiple feature maps (14) and determines combined feature maps (14); and a decoder network (16) that performs object recognition and / or semantic segmentation based on the determined combined feature maps (14).An output (19) of the decoder network (16) describes at least one object detected in the color image (10) and / or the infrared image (11) and / or an assignment of a class label to each pixel of the color image (10) and / or the infrared image (11).
Need to check novelty before this filing date? Find Prior Art
Citation Information
Patent Citations
Object detection and detection confidence suitable for autonomous driving
US20190258878A1
Electronic device and method for recognizing object by using plurality of sensors
US20200175714A1
Method and arrangement for determining a condition of a road surface by superposing images of NIR camera and RGB sensor
WO2020176029A1
End-to-end evaluation of perception systems for autonomous systems and applications
WO2022226238A1