OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Scenic

Scenic is an open-source Python research codebase built with JAX and Flax for developing, training and evaluating computer-vision models. It combines lightweight shared libraries for large-scale training tasks with project-specific model implementations and training and evaluation loops. Its documented work covers classification, segmentation and detection across images, video, audio and multimodal inputs.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Scenic — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Scenic when a vision-model research task fits an included project or baseline, or when shared JAX and Flax training components can support a custom experiment.

Who it is for

  • Computer-vision researchers developing or adapting model-training experiments.
  • Developers who want shared training utilities and project-specific pipelines in a JAX and Flax codebase.
  • Teams working on multi-device or multi-host vision-model training.

Documented uses

  • Launch experiments and manage summaries, logging and profiling.
  • Train and evaluate classification, segmentation and detection models.
  • Load and preprocess vision datasets through shared input pipelines.
  • Adapt a project through configuration changes or by customizing its data pipeline, model, losses, metrics or training loop.
Before you choose

Requirements & limits

The code is publicly available on GitHub under the Apache-2.0 license.

  • The core stack is Python, JAX and Flax.
  • Python 3.9 or later is required by the README quickstart.
  • The repository separates shared library-level utilities from project-specific code, allowing different levels of customization.
  • Project-specific dependencies may need to be installed separately.
  • The README documents a ViT-on-ImageNet training command and a minimal Colab for a simple feed-forward model.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up Scenic and start the documented ImageNet ViT run

Install Scenic from its repository and invoke the README’s ImageNet ViT training configuration.

Before you start

  • Python 3.9 or later
  • A local shell and access to the GitHub repository
  • Additional project-specific packages may be required; consult the baseline README or requirements file.

Steps

  1. Clone the Scenic repository with the command below.

    Documented command
    git clone https://github.com/google-research/scenic.git

  2. Change into the repository directory with cd scenic.

  3. Install the package with the command below.

    Documented command
    pip install .

  4. Invoke the documented run using the command below.

    Documented command
    python scenic/main.py -- --config=scenic/projects/baselines/configs/imagenet/imagenet_vit_config.py --workdir=./

Expected outcome

The documented command starts the ViT-on-ImageNet training workflow using the specified configuration and working directory.

Guide limitations

  • The README notes that specific projects and baselines may need extra packages listed in their README or requirements file.
  • The guide describes the repository’s documented setup and command, not an independently verified training result.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
google-research/scenic · Official homepage title · retained in its source language
How we review sourcesSuggest a correction