- This course is a deep dive into details of neural-network based deep learning methods for computer vision. During this course, students will learn to implement, train and debug their own neural networks and gain a detailed understanding of cutting-edge research in computer vision.
- Its course arrangement is similar to famous CS231n in Stanford, but this area develops super fast, and this course has newer lecture videos published in 2019.
- The repo is my implementation of six assignments in FA2020 courses.
- The assignments includes many topics and most of them focus on the implementation details about building blocks of the models instead of just throwing inputs into a blackbox to see how powerful deep learning models are.
- After finishing these assignments students can understand the implementation details in each model, such as how to implement back-propagation in FC, CNN and RNN without PyTorch.
- A1
- Basic tensor manipulation using PyTorch
- k nearest neighbors
- A2
- Linear Classifiers by SVM and Softmax
- Two layer neural network
- A3
- Fully Connected Network and Dropout
- Convolutional Neural Network and Batch Normalization
- A4
- Neural Network implementation using PyTorch
- Image Captioning through sequence models, such as vanilla RNN, LSTM and Attention
- Network Visualization. e.g. Saliency map, Adversarial Attacks and Class Visualization
- Style Transfer
- A5
- Single-Stage Object Detector using YOLO
- Two-Stage Object Detector using Faster R-CNN
- A6
- MNIST handritten digits generations by Variational Autoencoders
- MNIST handritten digits generations by Generative Adversarial Networks
- A1
- Most contents came from lecture slides, important concepts that lecturer mentioned in the video and some of my understandings.