Intro to Data Exploration

Prev Next

The video below introduces data exploration and search functionality using PASCAL VOC2012 public dataset. 


  1. Data exploration is available within a 'job.' The 'Jobs' page is reachable through Data Copilot > Data Visualization > Explore on the left navigation panel.

  2. The job visualization shows a plot panel and a thumbnail panel. The plot view gives a quick insight into the overall organization of data, with colors representing the different clusters.

  3. A point is selected by clicking on the point, resulting in the thumbnail panel populated with samples related to the point. Various sampling modes, like K nearest neighbors, group by cluster, etc., are available.

  4. Each thumbnail has controls to mark a point for similarity search, expand the image in full resolution, etc.

  5. A similarity search can be run on the entire image or a selected image region (patch search).

  6. The curated results through sampling or searching can be saved into a resultset for future retrieval and materialization (copying) into a chosen storage location like an S3 bucket or equivalent.