# google-research/uda

Unsupervised Data Augmentation (UDA)

Repository: https://github.com/google-research/uda
Canonical: https://ross.abutalabs.com/products/uda
Homepage: https://arxiv.org/abs/1904.12848
Language: Python
License: Apache-2.0
License Family: permissive
Topics: semi-supervised-learning, nlp, cv, tensorflow, computer-vision, natural-language-processing
Archived: true
Last push: 2021-08-28T07:16:56+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2633, "days_push": 1831, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2205, forks 313 (observed 2026-08-28T04:06:25.953120+00:00)

## What it is
Google Research's reference implementation of Unsupervised Data Augmentation (UDA), a semi-supervised learning method that uses advanced data augmentation like back-translation and RandAugment with consistency training. It includes code for BERT-based text classification, CIFAR-10/SVHN image classification, and back-translation augmentation, runnable on GPU and Cloud TPU.

## Use cases
- train text classifiers with very few labeled examples
- semi-supervised image classification on CIFAR-10 or SVHN
- reproduce UDA paper results
- back-translate text data for augmentation
- improve ImageNet accuracy with limited labels

## When to choose
- you have scarce labeled data and abundant unlabeled data
- you want to reproduce or build on the UDA research
- you need a proven semi-supervised baseline for NLP or vision

## When to avoid
- you need modern maintained tooling - it targets Python 2.7 and TensorFlow 1.13
- you want a production-ready training framework
- you lack GPU/TPU resources or large disk space for augmented data

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, nlp, image-processing
- domain: machine-learning, computer-vision
- platform: python
- tags: semi-supervised-learning, data-augmentation, tensorflow, bert, back-translation, research-code, natural-language-processing, research, gpu, linux

## Member repositories
- google-research/uda (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:25.953120+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:46:36.972530+00:00, confidence not recorded.
  - readme: https://github.com/google-research/uda (fetched 2026-08-28T04:06:25.953120+00:00, sha 17ef53a7df51)
  - homepage: https://arxiv.org/abs/1904.12848 (fetched 2026-08-29T10:27:00.576700+00:00, sha 805d2c5cbe4a)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T10:27:00.585749+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T10:27:00.589512+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T10:27:00.591648+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T10:27:00.587510+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.