视觉感知专家

该职位来源于猎聘 Major Responsibilities:

• Analyze customer and system requirements and translate them into vision perception algorithm, interface and performance requirements.
• Develop and optimize deep-learning and computer-vision algorithms for 2D/3D object detection, classification, tracking and attribute estimation for vehicles, pedestrians, cyclists and general obstacles.
• Develop road-structure perception including lane lines, road edges, traffic signs, traffic lights, road markings, stop lines, crosswalks and construction-area elements.
• Develop freespace, semantic segmentation, depth estimation, occupancy or BEV-based environment representation for driving and parking scenarios.
• Design and optimize temporal and multi-camera perception using front-view, surround-view or fisheye cameras; support camera-radar or camera-lidar fusion when required.
• Define data strategy and model-development workflows including data selection, annotation specifications, training, validation, hard-case mining and dataset quality improvement.
• Establish KPIs and evaluation methods for accuracy, recall, false positives, distance and velocity estimation, latency, resource consumption and robustness.
• Optimize models for embedded SoCs using quantization, pruning, operator optimization or hardware-aware deployment while meeting real-time and memory constraints.
• Lead robustness improvement for night, glare, tunnel transition, rain, snow, fog, occlusion, dirty lens, unusual objects and other long-tail scenarios.
• Support camera selection, installation assessment, intrinsic/extrinsic calibration, image-quality and ISP-related analysis together with sensor and hardware teams.
• Perform offline replay, SIL/HIL validation, vehicle testing, root-cause analysis and release verification; support customer reviews and production issue resolution.
• Prepare software requirements, algorithm design, test specifications and release documentation in line with Automotive SPICE, functional safety and SOTIF expectations. Profile: Required:
• Master's degree or above in computer science, artificial intelligence, electronic engineering, automation, robotics, automotive engineering or a related field.
• 5+ years of relevant experience in computer vision, deep learning or automotive perception algorithm development.
• Strong knowledge of modern vision perception methods, including object detection, semantic or instance segmentation, multi-object tracking, depth or 3D geometry and road-structure perception.
• Proficiency in Python and C/C++, and hands-on experience with PyTorch or TensorFlow and OpenCV.
• Experience in model training, dataset construction, evaluation, error analysis and systematic improvement of corner cases.
• Experience deploying and optimizing neural networks on embedded automotive SoCs or AI accelerators, with awareness of latency, memory and compute constraints.
• Experience with offline replay, SIL/HIL testing, vehicle validation and debugging in automotive or robotics environments.
• Strong analytical and problem-solving skills, ownership, teamwork and clear technical communication. Preferred:
• Experience with multi-camera perception, surround-view or fisheye vision for both driving and parking applications.
• Experience with BEV perception, occupancy networks, multi-view geometry, temporal fusion or transformer-based perception architectures.
• Experience with camera-radar/lidar fusion, environment modeling, visual localization, visual odometry or SLAM.
• Familiarity with camera calibration, image sensors, optics, ISP pipelines and image-quality evaluation.
• Experience with ADAS or automated-driving functions such as AEB, LKA, ACC, highway/city pilot, APA or RPA.
• Familiarity with Automotive SPICE, ISO 26262 and ISO 21448 (SOTIF).
• Good English communication skills for collaboration with global teams.