HW3 ― GAN, ACGAN and UDA

Overview

HW3 ― GAN, ACGAN and UDA

In this assignment, you are given datasets of human face and digit images. You will need to implement the models of both GAN and ACGAN for generating human face images, and the model of DANN for classifying digit images from different domains.

For more details, please click this link to view the slides of HW3.

Usage

To start working on this assignment, you should clone this repository into your local machine by using the following command.

git clone https://github.com/dlcv-spring-2019/hw3-
   
    .git

   

Note that you should replace with your own GitHub username.

Dataset

In the starter code of this repository, we have provided a shell script for downloading and extracting the dataset for this assignment. For Linux users, simply use the following command.

bash ./get_dataset.sh

The shell script will automatically download the dataset and store the data in a folder called hw3_data. Note that this command by default only works on Linux. If you are using other operating systems, you should download the dataset from this link and unzip the compressed file manually.

⚠️ IMPORTANT NOTE ⚠️
You should keep a copy of the dataset only in your local machine. DO NOT upload the dataset to this remote repository. If you extract the dataset manually, be sure to put them in a folder called hw3_data under the root directory of your local repository so that it will be included in the default .gitignore file.

Evaluation

To evaluate your UDA models in Problems 3 and 4, you can run the evaluation script provided in the starter code by using the following command.

python3 hw3_eval.py $1 $2
  • $1 is the path to your predicted results (e.g. hw3_data/digits/mnistm/test_pred.csv)
  • $2 is the path to the ground truth (e.g. hw3_data/digits/mnistm/test.csv)

Note that for hw3_eval.py to work, your predicted .csv files should have the same format as the ground truth files we provided in the dataset as shown below.

image_name label
00000.png 4
00001.png 3
00002.png 5
... ...

Submission Rules

Deadline

108/05/08 (Wed.) 01:00 AM

Late Submission Policy

You have a five-day delay quota for the whole semester. Once you have exceeded your quota, the credit of any late submission will be deducted by 30% each day.

Note that while it is possible to continue your work in this repository after the deadline, we will by default grade your last commit before the deadline specified above. If you wish to use your quota or submit an earlier version of your repository, please contact the TAs and let them know which commit to grade. For more information, please check out this post.

Academic Honesty

  • Taking any unfair advantages over other class members (or letting anyone do so) is strictly prohibited. Violating university policy would result in an F grade for this course (NOT negotiable).
  • If you refer to some parts of the public code, you are required to specify the references in your report (e.g. URL to GitHub repositories).
  • You are encouraged to discuss homework assignments with your fellow class members, but you must complete the assignment by yourself. TAs will compare the similarity of everyone’s submission. Any form of cheating or plagiarism will not be tolerated and will also result in an F grade for students with such misconduct.

Submission Format

Aside from your own Python scripts and model files, you should make sure that your submission includes at least the following files in the root directory of this repository:

  1. hw3_ .pdf
    The report of your homework assignment. Refer to the "Grading" section in the slides for what you should include in the report. Note that you should replace with your student ID, NOT your GitHub username.
  2. hw3_p1p2.sh
    The shell script file for running your GAN and ACGAN models. This script takes as input a folder and should output two images named fig1_2.jpg and fig2_2.jpg in the given folder.
  3. hw3_p3.sh
    The shell script file for running your DANN model. This script takes as input a folder containing testing images and a string indicating the target domain, and should output the predicted results in a .csv file.
  4. hw3_p4.sh
    The shell script file for running your improved UDA model. This script takes as input a folder containing testing images and a string indicating the target domain, and should output the predicted results in a .csv file.

We will run your code in the following manner:

bash ./hw3_p1p2.sh $1
bash ./hw3_p3.sh $2 $3 $4
bash ./hw3_p4.sh $2 $3 $4
  • $1 is the folder to which you should output your fig1_2.jpg and fig2_2.jpg.
  • $2 is the directory of testing images in the target domain (e.g. hw3_data/digits/mnistm/test).
  • $3 is a string that indicates the name of the target domain, which will be either mnistm, usps or svhn.
    • Note that you should run the model whose target domain corresponds with $3. For example, when $3 is mnistm, you should make your prediction using your "USPS→MNIST-M" model, NOT your "MNIST-M→SVHN" model.
  • $4 is the path to your output prediction file (e.g. hw3_data/digits/mnistm/test_pred.csv).

🆕 NOTE
For the sake of conformity, please use the python3 command to call your .py files in all your shell scripts. Do not use python or other aliases, otherwise your commands may fail in our autograding scripts.

Packages

Below is a list of packages you are allowed to import in this assignment:

python: 3.5+
tensorflow: 1.13
keras: 2.2+
torch: 1.0
h5py: 2.9.0
numpy: 1.16.2
pandas: 0.24.0
torchvision: 0.2.2
cv2, matplotlib, skimage, Pillow, scipy
The Python Standard Library

Note that using packages with different versions will very likely lead to compatibility issues, so make sure that you install the correct version if one is specified above. E-mail or ask the TAs first if you want to import other packages.

Remarks

  • If your model is larger than GitHub’s maximum capacity (100MB), you can upload your model to another cloud service (e.g. Dropbox). However, your shell script files should be able to download the model automatically. For a tutorial on how to do this using Dropbox, please click this link.
  • DO NOT hard code any path in your file or script, and the execution time of your testing code should not exceed an allowed maximum of 10 minutes.
  • If we fail to run your code due to not following the submission rules, you will receive 0 credit for this assignment.

Q&A

If you have any problems related to HW3, you may

You might also like...
Simple torch.nn.module implementation of Alias-Free-GAN style filter and resample
Simple torch.nn.module implementation of Alias-Free-GAN style filter and resample

Alias-Free-Torch Simple torch module implementation of Alias-Free GAN. This repository including Alias-Free GAN style lowpass sinc filter @filter.py A

Unofficial Alias-Free GAN implementation. Based on rosinality's version with expanded training and inference options.
Unofficial Alias-Free GAN implementation. Based on rosinality's version with expanded training and inference options.

Alias-Free GAN An unofficial version of Alias-Free Generative Adversarial Networks (https://arxiv.org/abs/2106.12423). This repository was heavily bas

GAN Image Generator and Characterwise Image Recognizer with python
GAN Image Generator and Characterwise Image Recognizer with python

MODEL SUMMARY 모델의 구조는 크게 6단계로 나뉩니다. STEP 0: Input Image Predict 할 이미지를 모델에 입력합니다. STEP 1: Make Black and White Image STEP 1 은 입력받은 이미지의 글자를 흑색으로, 배경을

Combine Tacotron2 and Hifi GAN to generate speech from text

EndToEndTextToSpeech Combine Tacotron2 and Hifi GAN to generate speech from text Download weights Hifi GAN - hifi_gan/checkpoint/ : pretrain 2.5M ste

Code of Adverse Weather Image Translation with Asymmetric and Uncertainty aware GAN
Code of Adverse Weather Image Translation with Asymmetric and Uncertainty aware GAN

Adverse Weather Image Translation with Asymmetric and Uncertainty-aware GAN (AU-GAN) Official Tensorflow implementation of Adverse Weather Image Trans

TransGAN: Two Transformers Can Make One Strong GAN
TransGAN: Two Transformers Can Make One Strong GAN

[Preprint] "TransGAN: Two Transformers Can Make One Strong GAN", Yifan Jiang, Shiyu Chang, Zhangyang Wang

Ultra-Data-Efficient GAN Training: Drawing A Lottery Ticket First, Then Training It Toughly
Ultra-Data-Efficient GAN Training: Drawing A Lottery Ticket First, Then Training It Toughly

Ultra-Data-Efficient GAN Training: Drawing A Lottery Ticket First, Then Training It Toughly Code for this paper Ultra-Data-Efficient GAN Tra

Implementation of 'lightweight' GAN, proposed in ICLR 2021, in Pytorch. High resolution image generations that can be trained within a day or two
Implementation of 'lightweight' GAN, proposed in ICLR 2021, in Pytorch. High resolution image generations that can be trained within a day or two

512x512 flowers after 12 hours of training, 1 gpu 256x256 flowers after 12 hours of training, 1 gpu Pizza 'Lightweight' GAN Implementation of 'lightwe

A collection of resources on GAN Inversion.

This repo is a collection of resources on GAN inversion, as a supplement for our survey

Owner
grassking100
A researcher study in bioinformatics and deep learning. To see other repositories: https://bitbucket.org/grassking100/?sort=-updated_on&privacy=public.
grassking100
FuseDream: Training-Free Text-to-Image Generationwith Improved CLIP+GAN Space OptimizationFuseDream: Training-Free Text-to-Image Generationwith Improved CLIP+GAN Space Optimization

FuseDream This repo contains code for our paper (paper link): FuseDream: Training-Free Text-to-Image Generation with Improved CLIP+GAN Space Optimizat

XCL 191 Dec 31, 2022
DR-GAN: Automatic Radial Distortion Rectification Using Conditional GAN in Real-Time

DR-GAN: Automatic Radial Distortion Rectification Using Conditional GAN in Real-Time Introduction This is official implementation for DR-GAN (IEEE TCS

Kang Liao 18 Dec 23, 2022
A Fast and Stable GAN for Small and High Resolution Imagesets - pytorch

A Fast and Stable GAN for Small and High Resolution Imagesets - pytorch The official pytorch implementation of the paper "Towards Faster and Stabilize

Bingchen Liu 455 Jan 8, 2023
Implementation of TransGanFormer, an all-attention GAN that combines the finding from the recent GanFormer and TransGan paper

TransGanFormer (wip) Implementation of TransGanFormer, an all-attention GAN that combines the finding from the recent GansFormer and TransGan paper. I

Phil Wang 146 Dec 6, 2022
Invert and perturb GAN images for test-time ensembling

GAN Ensembling Project Page | Paper | Bibtex Ensembling with Deep Generative Views. Lucy Chai, Jun-Yan Zhu, Eli Shechtman, Phillip Isola, Richard Zhan

Lucy Chai 93 Dec 8, 2022
Invert and perturb GAN images for test-time ensembling

Invert and perturb GAN images for test-time ensembling

Lucy Chai 49 May 2, 2021
HiFi-GAN: High Fidelity Denoising and Dereverberation Based on Speech Deep Features in Adversarial Networks

HiFiGAN Denoiser This is a Unofficial Pytorch implementation of the paper HiFi-GAN: High Fidelity Denoising and Dereverberation Based on Speech Deep F

Rishikesh (ऋषिकेश) 134 Dec 27, 2022
HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis

HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis Jungil Kong, Jaehyeon Kim, Jaekyoung Bae In our paper, we p

Rishikesh (ऋषिकेश) 31 Dec 8, 2022
Collection of generative models, e.g. GAN, VAE in Pytorch and Tensorflow.

Generative Models Collection of generative models, e.g. GAN, VAE in Pytorch and Tensorflow. Also present here are RBM and Helmholtz Machine. Note: Gen

Agustinus Kristiadi 7k Jan 2, 2023
GAN encoders in PyTorch that could match PGGAN, StyleGAN v1/v2, and BigGAN. Code also integrates the implementation of these GANs.

MTV-TSA: Adaptable GAN Encoders for Image Reconstruction via Multi-type Latent Vectors with Two-scale Attentions. This is the official code release fo

owl 37 Dec 24, 2022