|
所在平台: Coursera |
课程主页: https://www.coursera.org/learn/visual-perception-self-driving-cars
课程评论：没有评论
课程名称：自动驾驶汽车的视觉感知 课程概述： 欢迎参加“自动驾驶汽车的视觉感知”课程，这是多伦多大学自动驾驶汽车专业化的第三门课程。该课程将介绍自动驾驶中主要的感知任务，包括静态和动态物体检测，并调查常见的机器人感知计算机视觉方法。课程结束时，您将能够使用针孔相机模型，执行内外部相机标定，检测、描述和匹配图像特征，并设计自己的卷积神经网络。您将把这些方法应用于视觉里程计、物体检测与跟踪，以及可行驶表面的语义分割。这些技术构成了自动驾驶汽车感知系统的主要构建块。 课程的最终项目将要求您开发算法，以识别场景中物体的边界框，并定义可驾驶表面的边界。您将使用合成和真实图像数据，并在真实数据集上评估您的表现。 该课程为高级课程，适合具有计算机视觉和深度学习背景的学习者。成功完成此课程，您应具备Python 3.0的编程经验，以及线性代数（矩阵、向量、矩阵乘法、秩、特征值、特征向量和逆矩阵）方面的基础知识。 课程大纲： 1. 课程简介：介绍计算机视觉的基本概念，以及其在自动驾驶车辆感知方法中的应用，包括相机模型及其标定、单目和立体视觉、投影几何和卷积运算等。 2. 模块1：3D计算机视觉基础：继续探讨计算机视觉的主要概念，为自动驾驶车辆感知方法提供基础。 3. 模块2：视觉特征的检测、描述和匹配：介绍如何检测和跟踪特征，通过一系列图像进行运动跟踪，并在定位中与其他来源融合，还是深度网络中物体检测和语义分割的基础。 4. 模块3：前馈神经网络：简要介绍现代卷积神经网络的核心概念，强调在物体检测和语义分割等任务中有效的方法及构建和训练网络的基本架构与工具。 5. 模块4：2D物体检测：讲解深度神经网络在自动驾驶中的主要应用，包括物体检测和语义分割，为创建完整的自动驾驶感知管道提供基础技术。 6. 模块5：语义分割：深入探讨如何通过语义分割将图像像素与有用标签关联，以识别可驾驶表面，支持物体检测和车道边界评估。 7. 模块6：综合应用 - 可驾驶区域中的动态物体感知：该模块专注于实施碰撞警告系统，提醒自动驾驶汽车障碍物的位置和类别，项目包括：1) 3D可驾驶空间估计，2) 语义车道估计，3) 通过语义分割过滤物体检测的错误输出。
Name:Welcome to Course 3: Visual Perception for Self-Driving Cars
Description:This module introduces the main concepts from the broad and exciting field of computer vision needed to progress through perception methods for self-driving vehicles. The main components include camera models and their calibration, monocular and stereo vision, projective geometry, and convolution operations.
Name:Module 1: Basics of 3D Computer Vision
Description:This module introduces the main concepts from the broad field of computer vision needed to progress through perception methods for self-driving vehicles. The main components include camera models and their calibration, monocular and stereo vision, projective geometry, and convolution operations.
Name:Module 2: Visual Features - Detection, Description and Matching
Description:Visual features are used to track motion through an environment and to recognize places in a map. This module describes how features can be detected and tracked through a sequence of images and fused with other sources for localization as described in Course 2. Feature extraction is also fundamental to object detection and semantic segmentation in deep networks, and this module introduces some of the feature detection methods employed in that context as well.
Name:Module 3: Feedforward Neural Networks
Description:Deep learning is a core enabling technology for self-driving perception. This module briefly introduces the core concepts employed in modern convolutional neural networks, with an emphasis on methods that have been proven to be effective for tasks such as object detection and semantic segmentation. Basic network architectures, common components and helpful tools for constructing and training networks are described.
Name:Module 4: 2D Object Detection
Description:The two most prevalent applications of deep neural networks to self-driving are object detection, including pedestrian, cyclists and vehicles, and semantic segmentation, which associates image pixels with useful labels such as sign, light, curb, road, vehicle etc. This module presents baseline techniques for object detection and the following module introduce semantic segmentation, both of which can be used to create a complete self-driving car perception pipeline.
Name:Module 5: Semantic Segmentation
Description:The second most prevalent application of deep neural networks to self-driving is semantic segmentation, which associates image pixels with useful labels such as sign, light, curb, road, vehicle etc. The main use for segmentation is to identify the drivable surface, which aids in ground plane estimation, object detection and lane boundary assessment. Segmentation labels are also being directly integrated into object detection as pixel masks, for static objects such as signs, lights and lanes, and moving objects such cars, trucks, bicycles and pedestrians.
Name:Module 6: Putting it together - Perception of dynamic objects in the drivable region
Description:The final module of this course focuses on the implementation of a collision warning system that alerts a self-driving car about the position and category of obstacles present in their lane. The project is comprised of three major segments: 1) Estimating the drivable space in 3D, 2) Semantic Lane Estimation and 3) Filter wrong output from object detection using semantic segmentation.
Welcome to Visual Perception for Self-Driving Cars, the third course in University of Toronto’s Self-Driving Cars Specialization. This course will introduce you to the main perception tasks in autonomous driving, static and dynamic object detection, and will survey common computer vision methods for robotic perception. By the end of this course, you will be able to work with the pinhole camera model, perform intrinsic and extrinsic camera calibration, detect, describe and match image features and design your own convolutional neural networks. You'll apply these methods to visual odometry, object detection and tracking, and semantic segmentation for drivable surface estimation. These techniques represent the main building blocks of the perception system for self-driving cars. For the final project in this course, you will develop algorithms that identify bounding boxes for objects in the scene, and define the boundaries of the drivable surface. You'll work with synthetic and real image data, and evaluate your performance on a realistic dataset. This is an advanced course, intended for learners with a background in computer vision and deep learning. To succeed in this course, you should have programming experience in Python 3.0, and familiarity with Linear Algebra (matrices, vectors, matrix multiplication, rank, Eigenvalues and vectors and inverses).
