Responsibilities

1. Camera hardware selection, debugging, and calibration;

2. Design, develop, and optimize vision algorithms and models to solve image recognition, object detection, object tracking, and image segmentation problems;

3. Develop and advance deep learning-based monocular depth estimation and binocular stereo matching algorithm applications;

4. Research and advance 3D human pose estimation and 6D object pose estimation algorithms;

5. Track deep learning-based frontier visual SLAM algorithms;

6. Responsible for vision algorithm development, debugging, deployment, and implementation.

Requirements

1. Familiar with camera sensor working principles, ISP image processing pipelines, basic camera parameters and calibration, and image quality subjective/objective evaluation standards;

2. Familiar with mainstream neural network architectures (ResNet, ViT, etc.), object detection algorithms (YOLO, DETR, etc.), and semantic segmentation algorithms (Mask2Former, SAM, etc.);

3. Familiar with at least one common monocular depth estimation algorithm (e.g., DepthAnything) and at least one binocular stereo matching algorithm (e.g., LightStereo);

4. Understanding of mainstream 3D human pose estimation algorithms (e.g., PoseFormer) or 6D object pose estimation algorithms (e.g., FoundationPose) preferred;

5. Understanding of mainstream end-to-end visual SLAM algorithms (e.g., DROID-SLAM), and Gaussian Splatting-based SLAM (e.g., SplaTAM) preferred;

6. Proficient in PyTorch and other frameworks, familiar with common deep learning models, and master related implementation, training, tuning, and deployment;

7. Familiar with C++/C and Python programming languages, with good programming habits and engineering capabilities;

8. Full-time bachelor's degree or above in computer science, electronic engineering, or related fields; 2+ years of work experience.