This image was taken from State of the Art on Neural Rendering by Tewari et al.

–

Profolio

Here is the collection of final projects.

Title Authors link
Multiview Reconstruction with Neural Radiant Surfaces Jason Zhang, Sudeep Dasari Canvas
Disentangling Latent Space via VAE Jun Luo, link
Facial Expression Synthesis through AAE-CGAN Zheng Xu, link
Multi-objective unpaired image translationfor testing autonomous vehicles Trenton Tabor, Canvas
Towards Few-shot Image Synthesis Zhipeng Bao, Canbo Ye link
Attention-Based Image Translation George Cazanavette, Manuel Rodriguez Ladron De Guevara link
Multi-Human Pose to Image Synthesis Rawal Khirodkhar, Maneesh Bilalpur, Wentai Zhang link
Analysts of success and failure mode of existing single view synthesis methods Zihang Lai, link
Combining Inversion with Closed-Form Factorization for Image Manipulation Konwoo Kim, link
Improved Synthesis of Human Body and ClothTexture with Generative Adversarial Network Soyong Shin, Juyong Kim link
The Valley of Fear - Blending StyleGAN2 faces to cats Tarang Shah, Rohan Rao link
Creating implicit 3D representations from point clouds and/or RGB images Tarasha Khurana, Harsh Sharma, Divam Gupta link
Uncovering the facial expressions under the masks Chang Shi, Beilei Zhang, Sanil Pande link
Toon-Talker Clive Gomes, Rob Haber link
Memory Bank for Robust Image Generation Zijie Li, link
Conditional GAN with projection discriminator for image-to-image translation Jiaheng Hu, link
Interactive Texture Synthesis and ColorTransfer for Clothes Trung Nguyen, link
Controlling GAN Synthesis of Novel Architectural Imagery Michael Hasey, link
Photo-realistic Style Transfer Jingxiang Lin, link


Congratulations to all students for their amazing works!

Introduction

Welcome to the final project for the class. The purpose is to show us something novel based on the materials we cover in the class. You can try a new modification of a method, a particularly novel application, or a close analysis of the properties of an existing method. We’ll read over your project proposals and give feedback on them early on so that we can get awesome results on cool problems! Feel free to come to our office hours to discuss the progress and challenges over the rest of the semester. We’re happy to help!

Group Policy

You can work in groups of 1-3 people. We’ll expect the standard of work to be roughly proportional to the number of members in your group. In other words, larger groups will be graded to a somewhat higher standard as far as the scale of the project attempted and the amount of work completed.

Important Dates

  • 4/14: Project Proposal Due
  • 5/14, 1-3pm: Presentation Date
  • 5/17: Project Code and Website Due

Note: We will not allow late days on the project.

Project Proposal

We’d like to see a couple of paragraphs describing what you want to do for your project. Be sure to describe the end output, technique, novelty, dataset usage, and action plan. Include a couple of sentences placing your project proposal in context among related works. Submit this work as a pdf file to canvas. Feel free to include images or your hand-drawn figures. The page limit is two pages, but one page should be more than enough.

Presentation

You’ll need to give a 5-minute presentation about your project in class. We’ll announce the time for this soon, but you should give a quick presentation that offers an overview of the method and data and shows us the cool outputs of your work! If you can’t make the time we announce, we’ll ask you to submit an equivalent video.

Code and Website Submission

You’ll need to submit (1) the code for your project to canvas and (2) a website in the project directory of your website for the course as you did for other projects. This time, we’d really want to see through the description of the method, outputs of comparison methods (if applicable), the outputs of your algorithm, any math you do, and ablations if applicable. This will be the primary deliverable, and we encourage you all to do a good job with it, as you’ll be able to show people what you’ve made in a nicely presented way.

Good luck!