Skip to content

Latest commit

 

History

History
13 lines (10 loc) · 1.12 KB

README.md

File metadata and controls

13 lines (10 loc) · 1.12 KB

VQ-VAE for FIW

1. Introduction

This project is result of my learning process in the field of variational autoencoders. The dataset used for this use case is from project Families In the Wild, specifficaly from the task kinship verification. Competition, metrics and datasets are described in details in this paper.

VQ-VAE is central autoencoder architecture, and pixelCNN is used for prior distribution estimation. Models implementation and training procedure is higly inspired by homework solutions of Deep Unsupervised Learning course CS294-158-SP20.

2. Dataset

3. Code

4. Results

Reconstructions Figure 1: Reconstruction pairs, the first one is original and next one is reconstructed