DLCV Fall 2022

This repository contains my homework submissions for the "Deep Learning for Computer Vision" course at National Taiwan University (NTU), taught by Frank Wang.

Overview

HW1

1. Image Classification

• Train a CNN model from scratch
• Try alternative models/methods (e.g., fine-tune a pre-trained model)

2. Image Segmentation

• VGG16 + FCN32s (baseline model)
• Improved model

HW2

• Image Generation

  • Problem 1: GAN [face dataset - CelebA]
  • Problem 2: Diffusion models [digit dataset - MNIST-M]

• Unsupervised Domain Adaptation (UDA)

  • Problem 3: DANN [digit dataset - MNIST-M, SVHN and USPS]

HW3

• Problem 1: Zero-shot image classification with CLIP

• Problem 2: Image Captioning with Vision and Language Model

• Problem 3: Visualization of Attention in Image Captioning

HW4

• Problem 1: 3D Novel View Synthesis

• Problem 2: Self-supervised Pre-training for Image Classification