Computer Vision
计算机视觉研究使计算机从图像与视频中获取信息、理解视觉世界的理论与方法。课程内容包括图像处理基础、特征检测、相机模型与几何视觉、卷积神经网络、目标检测（YOLO、Faster R-CNN）、语义分割、人脸识别、3D视觉与生成模型（GAN、Diffusion）。学生将掌握视觉感知的核心算法，能应用于自动驾驶、医学影像、AR/VR等领域。
Computer Vision studies theories and methods that enable computers to extract information from images and video and understand the visual world. Topics include image processing fundamentals, feature detection, camera models and geometry, CNNs, object detection (YOLO, Faster R-CNN), semantic segmentation, face recognition, 3D vision, and generative models (GANs, Diffusion). Students master core perception algorithms applicable to autonomous driving, medical imaging, and AR/VR.
学生需基于YOLOv8实现一个实时目标检测系统。要求在自定义数据集上完成数据标注、模型微调、推理与性能评估，mAP@0.5需达到0.7以上。系统需支持视频流输入与实时可视化框选，可部署于本地或边缘设备。需提交代码、训练日志、演示视频及对比报告，讨论速度与精度的权衡。
Students implement a real-time object detection system based on YOLOv8. The project requires data annotation, model fine-tuning, inference, and performance evaluation on a custom dataset, achieving mAP@0.5 above 0.7. The system must support video stream input and real-time visualization, deployable on local or edge devices. Deliverables include code, training logs, a demo video, and a comparison report discussing the speed-accuracy trade-off.