课程名称：计算机视觉与模式识别

课程编号：X12PY0212，X12PY1220

课程层次：博士，硕士

任课教师：Ajmal Mian

开课学院：物理与光电工程

课程时间：2018年9月11日-9月21日

课程地点：西大楼419

该课程为引进的澳大利亚西澳大学计算机与软件工程学院的研究生优质课程，授课教师为西澳大学AjmalMian 副教授，主要涉及关于计算机视觉和模式识别的前沿研究内容。其主要包括：计算机视觉的深度学习基本原理、行为识别、人脸识别、3D重构、语义分割、高光谱图像超分、高光谱图像解混、高光谱图像分类等。该课程涉及的内容均为国际前沿研究问题，国外的研究水平高于国内，有助于开阔我校研究生的国际视野、学到世界前沿知识。

课程内容介绍及实施计划：

Day – 1 (Computer Vision)

1 Image analysis (Introduction and image formation, convolution and filtering)

2 Convolution and image filtering

Day – 2 (Computer Vision)

3 Feature extraction (lecture 7: HAAR, HOG, LBP, SIFT)

4 Image classification

Day – 3 (Computer Vision)

5 3D computer vision

6 3D object recognition and segmentation

Day – 4 (Remote Sensing)

7 Hyperspectral imaging, classification, unmixing

8 Hyperspectral super resolution

Day – 5 (Deep Learning)

9 Multilayer perceptron

10 Convolutional Neural Networks

Day – 6 (Deep Learning)

11 Image classification with CNNs (AlexNet, VGG)

12 Transfer Learning (pre-trained CNNs, fine-tuned CNNs, re-trained CNNs)

Day – 7 (Deep Learning)

13 Famous CNN architectures (ResNet, Inception, Inception ResNet)

14 Semantic segmentation with CNN (hypercolumns cvpr15, fully convolutional network cvpr16)

Day – 8 (Deep Learning)

15 Deep learning from videos (two stream vgg, C3D, C2-3D, LSTM, FTP)

16 Data augmentation and synthesis (Blender, playing for data etc)

Day – 9 (Deep Learning)

17 Deep learning from 3D data (depth images, normal vectors)

18 Deep learning from point clouds (classification + segmentation)

Day – 10 (Deep Learning)

19 Deep learning from unconventional image/non-image data (Light field, hyperspectral, marker)

       20 Concluding remarks, libraries for deep learning etc.