OPEN SOURCE SOFTWARE
Some of our products are available for anyone to use.
LeoFS is a distributed object storage which offers highly available, distributed, eventually consistent storage system. It suppors S3-API, Multi Data Center Replication, Built-in the Cache Mechanism and so on. It's possible to build in-house fast object storage in stead of S3 and to deal with secure files. LeoFS will make a great contribution to reduce cost.
Egison is the pattern-matching-oriented, pure functional programming language. Egison simplifies code by replacing complex nested loops and conditional branches with an intuitive pattern-matching expression. Egison is improving day by day. Recently, we added the facility to handle databases with intuitive Egison pattern-matching. We also released Ruby extension for Egison pattern-matching.
ROMA is a distributed key-value store implemented by Ruby. It offers high availability and scalability with pure P2P architechture. The most significant feature of ROMA is flexibility to add customable functions on ROMA developed by engineers. You can use not only simple string but also list on ROMA. Furthermore, ROMA project also provides various of clients such as Ruby, PHP, and Java.
fairy is a framework for distributed processing like MapReduce in Ruby. fairy supports programming model, called filter IF, and various built-in filters. It offers high productivity and simplicity because of its API like Ruby programming. It's expected to increase productivity.
Category2Vec is an implementation of the category vector models [Marui and Hagiwara 2015], and the paragraph vector models [Le and Mikolov 2014]. These programs are based on word2vec [Mikolov et al. 2013a,b] in gensim project [Rahurek 2013]. After training, you can obtain distributed representations for categories, paragraphs, and words. You can also infer a category from a description.
ByNet-SR is the implementation of our ICIP2017 paper: “ByNet-SR: Image Super Resolution with a Bypass Connection Network”. The model takes a low-resolution image as the input, and outputs a sharp high-resolution image. The evaluation in the paper shows that it is currently one of the best published methods for image super-resolution. Code for both training and inference is provided. Our implementation uses the PyTorch framework, and can be run in GPU or CPU mode.