实验课程

人工智能硬件综合实践

本课程旨在介绍部署驱动的人工智能系统设计方法，从基础模型和应用场景出发，要求学生掌握深度学习的技术、术语和数学等背景知识，了解深度神经网络基本架构与图像分类、目标检测、人体姿态估计、典型下游任务，学习如何恰当地构建和训练这些模型达到最佳结果；同时结合典型边缘部署场景，进一步探索自适应模型压缩、结构优化、数据增强、硬件加速等前沿技术的实施方法，让学生在充分理解与掌握基础知识的同时，也能尝试业界最前沿的系统设计空间探索，深入体会理论知识如何落地并指导实际产业界开发。本课程通过平时作业与小组大作业相结合的方式进行考核，训练学生在开发设计复杂人工智能系统时的独立思考和分工协作能力。

This course aims to introduce deployment-driven artificial intelligence system design methods. Starting from basic models and application scenarios, students are required to master the background knowledge of deep learning technology, terminology and mathematics, understand the basic architecture of deep neural networks and typical downstream tasks, such as image classification, target detection, and human pose estimation, learn how to properly build and train these models to achieve the best results. At the same time, combined with typical edge deployment scenarios, further explore the implementation methods of edge-end technologies such as adaptive model compression, structure optimization, data enhancement, hardware acceleration, etc., so that students can fully understand and master the basic knowledge, but also can try the most advanced Design Space Exploration methodology in the industry, which helps in-depth understanding of how theoretical knowledge is implemented and promote the development of the actual industry. This course is assessed through a combination of regular homework and grouped homework, and trains students to think independently and to work together in the development and design of complex artificial intelligence systems.