Model Create

The fastest way to build and train AI models

Sign up now for free

Simple tools to create, combine and update AI in the enterprise 

Save valuable time and resources with the Create API and user-friendly tools that make it easy for your business to develop custom AI applications.

Our advanced knowledge transfer algorithms and neural network architecture ensure highly accurate and performant results - even if you have no background in machine learning and minimal training data.

retail-store-layout-planning

Deep learning is complex. How does Clarifai make
creating new models so easy?

At Clarifai, a model is born the moment you click “Train”

 

Create a concept

Screen Shot 2020-02-03 at 6.02.32 PM

Label your data

label_breakfast

Click “Train”

train

 


Build your own AI

Clarifai provides many different models that 'see' the world differently. Each model contains a group of concepts, and can only recognize the concepts it contains.

There are times when you wish you had a model that sees the world the way you see it. Clarifai helps you make this wish come true. You can create a model and train it with your own images and concepts. Once trained, you can use the model to make predictions on new data.

Base Model CNNs

Clarifai uses convolutional neural networks (CNN) for visual recognition. Our collection of base models are constantly re-trained on high-performance machines using millions of labeled inputs, over a span of weeks.

Users can train powerful models on top of our base models in seconds, instead of weeks. It only takes a few inputs to get started.

General

Travel

Food

Wedding

Moderation

Furniture

Textures

Art styles

Vehicles

Dog breeds

Apparel

Real Estate

Hair style

Roof type

Model versions

If you’ve ever developed software, you know the importance of version control. We bring this practice to AI by creating a new model version every time you train your model.

This ensures that you can go back to a previous version if you make a mistake, or if the model performance degrades. Model versions also make it easier to scale your AI efforts through collaboration. You can have one model version in use in your deployed products, while other teams are hard at work building the next iteration of the model.

model_versions-1

Why do models need to be “trained”?

Machines learn by connecting the dots between annotations and input data.

This is just like teaching a child to recognize something new. Give a child a few examples of what you want them to learn, and once they understand the pattern they can recognize new examples.

When you click “Train” Clarifai builds or modifies a custom model by looking at your examples and learning how to recognize your annotations.

mission_style

The tools you need to fine-tune the accuracy of your custom models

Take the guesswork out of evaluating the performance of your custom models. We automate the testing and validation process by generating random subsets of your training data to use as test data and then performing a “5-split cross validation”.

Images

Ensure that you have supplied enough of the right example images

versions

Evaluate and showcase all the versions of the model you have trained

test

Predict against your own designs and test performance before using in a production environment

adjust

Adjust thresholds for recall and precision rates to optimize for accuracy

Understanding 5-split cross validation

Train-test splits give you an indication of how your model will perform on unseen data.

  • We split your training data into five parts. Then we set aside one part to use as a test set and use the remaining data to train a new model.
  • Once that model is created we make predictions against the test set using the new model.
  • Then we compare these predictions against the labels given for the inputs.
  • This process is repeated four more times and then each concept is given a score.

Review the training data used by each of your model versions. If your predictions and your labels aren't matching up, you can easily make adjustments by using the labelling functionality built right into our APIs and Clarifai Portal.

eval-metrics-1.0@2x-1

Clarifai partners with organizations of all sizes to deploy cutting-edge artificial intelligence

World leading AI

  • Recognized as an industry leader by Forrester
  • World-class research scientists named on over 1250 publications drawing over 140K citations
  • Incredible accuracy from day one using our dozens of pre-trained models
  • Knowledge graph linking to understand concepts in multiple languages
  • GDPR and Privacy Shield compliant

Up and running in minutes

  • Get started with 5,000 free operations each month, no credit card required
  • The fastest UX to add your inputs and train your models
  • Up-to-date developer documentation and customer-first support
  • API clients in 6 programming languages
  • RESTful and gRPC based API for wide adoption and high performance

Scaleable for your business

  • Cloud, on-premise, or air-gapped: We deploy to fit your needs
  • Fine-grained permissions let your organization collaborate
  • Models get smarter and more accurate as data flows into our API
  • Fast responses enable real-time applications
  • Output auto-scales to any incoming bandwidth

Built for developers from the ground up

Clarifai was founded by developers and designed to help developers start building immediately and quickly master AI

curl -X POST \
-H "Authorization: Key YOUR_API_KEY" \
-H "Content-Type: application/json" \
https://api.clarifai.com/v2/models/<model_id>/versions
app.models.train("{model_id}").then(
function(response) {
// do something with response
},
function(err) {
// there was an error
}
);
// or if you have an instance of a model
model.train().then(
function(response) {
// do something with response
},
function(err) {
// there was an error
}
);
from clarifai.rest import ClarifaiApp
app = ClarifaiApp(api_key='YOUR_API_KEY')
model = app.models.get('{model_id}')
model.train()
client.trainModel("{model_id}").executeSync();
ClarifaiImage *image = [[ClarifaiImage alloc] initWithURL:@"https://samples.clarifai.com/puppy.jpg"]
[app getModel:@"{id}" completion:^(ClarifaiModel *model, NSError *error) {
[model train:^(ClarifaiModel *model, NSError *error) {
NSLog(@"model: %@", model);
}];
}];

Model pricing summary

Pay as you go. Our prices for building, training and hosting models scale with you as your business grows


Request
Free tier limit operations First 500,000 operations Next 4.5 million operations Next 45 million operations Over 50 million operations
  (all prices quoted per 1,000 operations and tiers calculated monthly)
Add inputs 5,000 $1.20 $1.10 $1.00 $0.90
Model train 5,000 $1.20 $1.20 $1.20 $1.20




 


Request
Free tier limit
inputs
First 500,000
inputs
Next 4.5 million
inputs
Next 45 million
inputs
Over 50 million
inputs
  (all prices quoted per 1,000 operations and tiers calculated monthly)
Custom model hosting 10 $0.50 $0.50 $0.50 $0.50

 

Committed use discounts

Commit to a monthly volume and receive a significant discount beyond standard volume discounts. Contact our sales team to learn more.

Talk to sales