OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

KerasHub

KerasHub is a Python library that provides Keras 3 implementations of model architectures paired with pretrained checkpoints, exposed through Keras layers and models.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

KerasHub — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use KerasHub when you want model architectures and pretrained checkpoints within the Keras 3 API, or want to fine-tune a supported model using JAX, TensorFlow or PyTorch.

Who it is for

  • Keras developers choosing pretrained models for an application or experiment.
  • Teams building text, image or audio generation and classification workflows.
  • Practitioners fine-tuning models with JAX, TensorFlow or PyTorch.

Documented uses

  • Browse pretrained model presets for a Keras 3 project.
  • Run image classification with a pretrained ResNet model.
  • Fine-tune a pretrained BERT classifier on IMDb movie reviews.
  • Build a prototype around a pretrained model for text, image or audio tasks.
Before you choose

Requirements & limits

  • KerasHub requires Keras 3; the README specifies Python 3.11 or newer.
  • Choose and set the backend before importing Keras libraries.
  • KerasHub installation pulls in TensorFlow for tf.data preprocessing, even when training uses another backend.
  • Underlying third-party pretrained models may have separate licenses from the library’s Apache-2.0 license.
  • The README cautions that APIs may change during pre-release 0.y.z development.

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

Install KerasHub and select a backend

Prepare a KerasHub environment with one of its documented backends.

Before you start

  • Python 3.11 or newer, as indicated by the repository badge
  • Keras 3

Steps

  1. Install or upgrade KerasHub with the command below.

    Documented command
    pip install --upgrade keras-hub

  2. Set KERAS_BACKEND to jax, tensorflow or torch before importing Keras libraries.

Expected outcome

KerasHub is installed and configured to use the selected Keras backend.

Guide limitations

  • Installation also pulls in TensorFlow for the tf.data preprocessing API.
  • The README warns that APIs may change during pre-release 0.y.z development.
Documentation 1 ↗
Source-based practical guide

Use a pretrained model for image classification

Load the documented ResNet preset and request a prediction for an image.

Before you start

  • Keras 3 and KerasHub installed
  • A configured JAX, TensorFlow or PyTorch backend
  • NumPy and an image accessible to Keras image utilities

Steps

  1. Create an ImageClassifier from the resnet_50_imagenet preset with softmax activation.

  2. Load an image using Keras image utilities and convert it to a NumPy array.

  3. Call predict on a batch containing the image and decode the ImageNet predictions.

Expected outcome

The classifier returns predictions that can be decoded as ImageNet labels.

Guide limitations

  • The example uses the ImageNet ResNet preset; the README does not establish that every checkpoint is suitable for every image-classification task.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
keras-team/keras-hub · Official homepage title · retained in its source language
How we review sourcesSuggest a correction