Keynotes 大師開講

112學年暑期課程：視覺大數據的深度學習

黃正能教授：Prof. Jenq-Neng Hwang (IEEE Fellow) of the University of Washington, Seattle, USA

一、 中文課程名稱：視覺大數據的深度學習 

二、英文課程名稱：Deep Learning for Big Visual Data

三、 實際授課期間(暑期上課日期為7/1-8/30，上課須至少為6週)：

四、 授課時間：每週 一，第_5___節至第__8__節; 7/15, 7/22, 7/29, 8/5, 8/12

                             每週 四,   第_5___節至第__8__節; 7/11, 7/18, 7/25, 8/1, 

Topics:

1. Multilayer Perceptrons and Backpropagation Learning (course introduction, training loss functions, performance metrics, neural networks structures and learning mechanisms)

2. Convolution Neural Networks (from MLPs to CNNs, various CNNs for image classifications, transfer learning of CNNs)

3. Practical Usage of CNNs (few shot learning, metric learning, face identification and verification, long tailed recognition, unsupervised domain adaptation) 

4. Image Object Detection and Multi-Object Tracking (two-stage and one-stage detector, faster RCNN, Yolo, CenterNet, Tracking by Detection, Hungarian Assignment)

5. Image Segmentation and Human Pose Estimation (semantic segmentation, instance segmentation, 2D human pose estimation, 3D pose estimation)

6. Transformers for Large Language Models and Visual Applications (self-attention, transformer encoder and decoder, BERT, decoder based large language models, ChatGPT, finetuning of LLM, Vision transformer, Detection Transformer)

7. Generative Adversarial Networks (generator and discriminator, conditional GAN, style GAN, pixel2pixel GAN, CycleGAN)

8. Diffusion Models for Image/Video Generation (DDPM, Stable Latent Diffusion,Dall-E, Cascaded Diffusion, Video diffusion model)

Remark: 4 hours on each topic, and the last 4 hours for students' report presentations!