Skip to content

CS492(A): Machine Learning for 3D Data

Minhyuk Sung, KAIST, Spring 2022

Project Proposal

Submission Deadline: Mar 30 (Wed), 23:59 KST
Where to submit: KLMS

What to Submit

Submit a write-up (up to two pages) as a PDF file.
Including the followings:

Research Track
  • Team member names
  • Project title
  • Problem definition
  • Previous work
  • Novelty and technical contributions of the proposed idea
  • Key technical ideas
  • Experiment plans including datasets, evaluation metrics, and baseline methods
  • Timeline (Be specific.)
    • 1st midterm check-in plans (Toy experiments)
    • 2nd midterm check-in plans
  • Role of each team member
Development Track
  • Team member names
  • Paper title and link
  • Problem definition
  • Key technical ideas
  • Experiment plans including datasets, evaluation metrics, and baseline methods
  • Expected challenges in the implementation
  • (Optional) The components you plan to use existing code
  • Timeline (Be specific.)
    • 1st midterm check-in plans (Toy experiments)
    • 2nd midterm check-in plans
  • Role of each team member

Midterm Check-Ins

There will be two midterm check-ins, and the proposal needs to include specific plans for each midterm check-in. You will need to submit a short write-up for each midterm check-in and may get a penalty if you do not accomplish the tasks you planned in the proposal.

  • 1st midterm check-in: Due May 2 (Mon)
  • 2nd midterm check-in: Due May 16 (Mon)

The plans for the 1st midterm check-in must include toy experiment plans:

Research Track
  • Implement baseline methods (that will be compared with your method) and report the results.
Development Track
  • Run the authors’ code and report the results. Check whether the outputs match the results in the paper.
  • Try to run the network training code instead of just using a pretrained model. If the results in the paper are not reproducible using the authors' training code, you may need to reconsider the project idea. Consult the instructor or TAs ASAP.
  • Note that you are NOT allowed to use the authors' code in your implementation, but it is still strongly recommended to check out the authors' code and see whether it works as described in the paper.

Penalty

  • Late submission: 10% penalty in the project evaluation.
  • Missing components in the write-up: 5% penalty in the project evaluation.