CSE252D Spring 2022 HW1

1. Submission Instructions

1. Attempt all questions.
2. Please comment all your code adequately.
3. Include all relevant information such as text answers, output images in notebook.
4. Academic integrity: The homework must be completed individually.
5. (Imp) Correctly select pages for each answer on Gradescope to allow smooth grading.
6. Due date: Assignments are due Thu Apr 28, 11:59PM PDT.
7. Access this assignment by cloning this repository using,
   git clone https://github.com/ViLab-UCSD/cse252d-sp22-hw1.git
   Note the changes in Github policy for cloning with personal access tokens here.
8. The entry point to the assignment is the Jupyter Notebook hw1-CSE252D.ipynb
9. Follow the rest of README (this file) for instructions on how to setup your environment, data and compute.
10. Submit the PDF version of your notebook and your code on Gradescope.
    (1) Convert the ipynb file to pdf and upload it to Homework 1 writeup. Select pages for each answer.
    (2) Compress your notebook, code and supporting results to zip and upload it to Homework 1 code. Do not include any dataset or large data files.
11. Rename your submission files as Lastname_Firstname.pdf and Lastname_Firstname.zip.

2. Setup Jupyter

2.1. [Option 1] On your own machine

• Install SWIG
  • On Ubuntu: sudo apt-get install swig (sudo required)
  • On MacOS: brew install swig
    You need to install Homebrew first with HomeBrew
• Install Python 3.X and Pip
• [Recommended] Create an environment (e.g. with Anaconda)
  conda create --name py36 python=3.6 pip
conda activate py36
• Install Jupyter Notebook
  conda install jupyter
• Install kernels for Jupter Notebook
  conda install nb_conda
• Launch Jupyter Notebook server
  jupyter notebook
  You will be provided with a URL that you can open locally.
  In an opened notebook, change the kernel (Menu: Kernel -> Change Kernel) to the name of the conda env you just created (in this case py36).

2.2. [Option 2 Recommended] On Data Science & Machine Learning Platform

• (local) (IMPORTANT) Connect to UCSD VPN
• (local) Login with your Active Directory credentials
  ssh {USERNAME}@dsmlp-login.ucsd.edu
• Launch your pod.
  • You should enter a node with 1 GPU, 8 CPU, 16 GB RAM, with normal priority (running up to 6 hours
    • launch-scipy-ml.sh -i ucsdets/cse152-252-notebook:latest -g 1 -m 16 -c 8 -p normal
  • To enable longer runtime k (up to 12) hours with normal priority
    • K8S_TIMEOUT_SECONDS=$((3600*k)) launch-scipy-ml.sh -i ucsdets/cse152-252-notebook:latest -g 1 -m 16 -c 8 -p normal
  • To enable longer runtime k (more than 12) hours with lower priority
    • K8S_TIMEOUT_SECONDS=$((3600*k)) launch-scipy-ml.sh -i ucsdets/cse152-252-notebook:latest -g 1 -m 16 -c 8
  • To run your container in the background up to 12 hours, add -b to above command. See details here.
• You will be provided with a URL that you can open locally:
  Click on the link and navigate to the Jupyter notebook.
• If you cannot launch a pod, set up the environment following these instructions.
• You can also launch a pod by logging in at https://datahub.ucsd.edu/hub and selecting the right environment:

3. Setup library dependencies

• Having created a conda environment, clone this repository using
  git clone https://github.com/ViLab-UCSD/cse252d-sp22-hw1.git
cd cse252d-sp22-hw1
• Install dependencies using pip
  pip install -r requirements.txt --user
• Install pyviso using
  cd pyviso/src/
pip install -e . --user

4. Access Data

On the dsmlp.ucsd.edu server, the datasets are located at

• Q1: SfM
  /datasets/cs252d-sp22-a00-public/dataset_SfM
  Change the dataset path in jupyter notebooks accordingly.
• Q5:
  /datasets/cs252d-sp22-a00-public/sfmlearner_h128w416
/datasets/cs252d-sp22-a00-public/kitti
• You can also access the data on Canvas under Files.

5. How to run

• Q1-Q4: SfM - Working folder: ./pyviso
• Launch Jupyter Notebook
  There is a hw1-CSE252D.ipynb jupyter notebook file in the top-level directory cse252d-sp22-hw1.
• Options
  One can toggle if_vis = True/False allows you to enable/disable the visualization.
• Output
  The errors are printed and the visualizations are saved at vis/. The images should look like:
• To fetch the files you can use scp to transfer files from the cluster to your local machine: scp -r <USERNAME>@dsmlp-login.ucsd.edu:<PATH TO THIS REPO>/pyviso2/vis {LOCAL PATH}
• [Extra] You can run your container in backgound using TMUX.