Given a grayscale photograph as input, this paper attacks the problem of hallucinating a plausible color version of the photograph. This problem is clearly underconstrained, so previous approaches have either relied on significant user interaction or resulted in desaturated colorizations. We propose a fully automatic approach that produces vibrant and realistic colorizations. We embrace the underlying uncertainty of the problem by posing it as a classification task and use class-rebalancing at training time to increase the diversity of colors in the result.

Julia is a high-level, high-performance dynamic programming language for numerical computing.

A tool that launches docker jobs on Amazon EC2 cloud instances (machines).

This repository contains the authors' implementation from "Data Augmentation using Learned Transformations for One-shot Medical Image Segmentation", which will be presented as an oral at CVPR 2019. We provide code for training spatial and appearance transform models, and for using the transform models to synthesize training examples for segmentation.

We present Carnelian, a pipeline for alignment-free functional binning and abundance estimation, that leverages low-density even-coverage locality sensititve hashing to represent metagenomic reads in a low-dimensional manifold. When coupled with one-against-all classifiers, our tool bins whole metagenomic sequencing reads by molecular function encoded in their gene content at significantly higher accuracy than existing methods, especially for novel proteins.

Electronic Medical Record Phenotyping using the Anchor and Learn Framework.

Metagenomic binning using low-density hashing a support vector machine

MICA (Metagenomic Inquiry Compressive Acceleration) is a family of programs for performing compressively-accelerated metagenomic sequence searches based on BLASTX and DIAMOND. MICA also includes compressively accelerated versions of the BLASTP family of tools (including PSI-BLAST and DELTA-BLAST), as well as a compression tool (mica-compress) for creating searchable, compressed databases based on an input FASTA file.

While it is nearly effortless for humans to quickly assess the perceptual similarity between two images, the underlying processes are thought to be quite complex. Despite this, the most widely used perceptual metrics today, such as PSNR and SSIM, are simple, shallow functions, and fail to account for many nuances of human perception. Recently, the deep learning community has found that features of the VGG network trained on ImageNet classification has been remarkably useful as a training loss for image synthesis. But how perceptual are these so-called "perceptual losses"?

Torch implementation for learning a mapping from input images to output images.