Vision Copilot for data science teams

Connect

No ingestion or copying data

Access your visual data and metadata from any cloud provider (S3, Azure Hot, Google Cloud Storage) or from a local source without worrying about moving or copying it outside the your environment.

Explore

Identify latent structures of visual datasets

Interactively explore large volumes of visual datasets through a choice of clustering and embedding techniques, identify outliers, sample, detect biases and filter content.

Search

Image-based search

Perform interactive visual search across millions of images or large video datasets and intuitively refine to build effective training datasets.

Patch Search

Granular Visual Search

Perform fine grainded search by simply drawing a region of interest "Patch" on example image or video frames.

Text to Image Search

Query images with free-form text

Perform search based on natural language (NLP) on large image and video datasets without any annotations. Find relevant objects faster.

Video Sequence Search

Search for Scenarios

Find interesting events or video sequences in large video dataset. Focus your search by interactively picking key frames.

Analyze

Evaluate model accuracy

View model performance from multiple lenses and an aggregate view to identify the causes of inaccuracies within your model by isolating and focusing on the data that matters.

Saliency Map

See what your Model sees

Easily measure the spatial support of a particular class of an image with regards to your computer vision models. See Salient features of an image that influences your model’s inference. Understand why model predictions are incorrect.

Visualize

Visualize Annotations and Image Attributes

Visualize existing data attributes for data selection, curation or model accuracy analysis.

Generate and visualize new metadata info in the form of Classification, Object detection or Segmentation with models of your choice.