Machine Learning Workflows

Use Mesh Workflows to easily process complex logic and gain insights.

With Mesh Workflows, you can work with your machine learning models like they are nodes in a graph. Clarifai makes it easy to connect together models of different kinds so that you can perform complex operations on your data, and build solutions that target your specific business needs. 

• 

  Deploy Instantly

  No dev-ops required. Deploy with the click of a button. 

• 

  Update Your Models

  Iterate easily and build upon your models with model versioning tools.

• 

  Optimize Performance

  Speed up your application performance with optimized parallel processing.

• 

  Index Your Data Automatically

  Mesh Workflows ensure that your data is indexed and searchable.

Mesh Workflows work with different input and output types flexibly

The nodes in your Mesh Workflows can accept data from your database or other models. Just select a model that accepts the type of input that you would like to work with.

• Images

  Images can be processed by machine learning models, analyzed, cropped, and used to power search. 

• Concepts

  The concepts detected by your model can be used as inputs. This can help you to label and structure your data. 

• Embeddings

  The abstract mathematical representation generated by your machine learning model can be used to power new custom models. 

• Regions

  Many models can accept specific regions of an image as an input, helping you understand the specific objects in your visual data.

• Regions with Concepts

  Models can accept specific regions of an image as an input along with any detected concepts that your model has identified.

• Text

  Text can be used as an input into natural language processing (NLP) models. 

• Regions with Text

  Detect text in images and then prepare this data for analysis by natural language processing (NLP) models.

• Video Frames

  Individual frames of video can be used as inputs much like individual images can be used as inputs. 

• Concepts

  Models can be trained to recognize concepts in your video, text and image data. 

• Embeddings

  The abstract mathematical representation generated by your machine learning model can be used to power new custom models or search. 

• Regions with Embeddings 

  Models can provide embedding representations of specific regions of your images and videos so that you can learn more about specific objects in your data. 

• Regions with Text

  Detect text in images and then prepare this data for analysis by natural language processing (NLP) models.

• Text

  Text is a common output for many object character recognition (OCR) and natural language processing (NLP) models. 

• Regions with Concepts

  Detect unique objects in your images and learn about what the object is. 

• Clusters

  Clusters are groups of semantically similar images and video frames. Clusters help you to search your data based on visual similarity. 

• Colors

  Detected colors in an image can be used to search, sort and filter data.

Mesh Workflows help you create, improve and discover with AI

Learn about some of the ways that you can use machine learning models and Mesh Workflows to power your business with the fastest time-to-accuracy in AI. 

• Action triggers

  React to the real-world faster that ever was possible before with the power of AI. Mesh Workflows can turn the predictions made by your AI models into action. Send alerts, trigger external systems, or even route data for further processing. 

• Multi-modal AI

  Mesh lets you use one platform to process many different types of data. Build inference pipelines that recognize text in images (OCR), and understand that text with natural language processing (NLP). Chain together multiple transformations easily for deeper insights. 

• Active learning

  Continuously improve your production models with auto-labeling, and keep a “human in the loop” to ensure model quality. Mesh allows to build your own custom training loop.

• Noise reduction

  Focus your dataset with helpful prebuilt models that can quickly clean up many of the most common use cases. Reduce the noise caused by unwanted data so that you can build more efficient and accurate models. 

• Domain filtering

  Build separate inference pipelines for your data based on high level categories in your dataset. Parse a large and complex dataset into manageable chunks. Get to the data you need quickly. 

• Complex classification

  Learn about the many dimensions of your dataset by chaining together multiple classification models. Learn about multiple categories of information at once and optimize each model individually. 

• Object identification

  Learn about what is “in” your visual data with automatic object identification. Detection models can even detect objects in specific regions of an image and then pass this region to other models for further processing. 

• Object-to-object search

  Search for like objects in your database. Perform searches based on visual similarity alone. Mesh workflows can classify and detect images that power your search queries.