ToolZ

Ever had to annotate a large dataset, in excess of 10k images per class? Until your eyes bleed? We have. So much so, that we created a product to automate what could be automated, and make it easier for what could not.

ToolZ eases the pain of annotating images by leveraging the extraordinary speed of Metal and the Neural Engine of your iPad. Annotating from your couch is a feature. We are confident that once you have spent five minutes creating boxes using the Apple Pencil™️, you won’t go back to your mouse and keyboard. (The Apple Pencil is not a requirement, we just find it useful for the accurate placement of boxes).

Features

  • Annotate images for classification or object detection
  • Use your Apple Pencil to navigate between images, objects and classes
  • Work with semi-automated annotation; use a model to generate bounding boxes, transfer them to annotations, quickly review them on a collection view
  • Filter images by colour, contrast and size ranges
  • Create configurations for different datasets
  • Run CoreML models on large sets of images, either on the CPU, GPU or Neural Engine
  • Import and export images, zip or tar files
  • Solves the unstructured data problem; ToolZ writes annotations into the EXIF metadata of the images
  • Export to CoreML YOLO JSON, Pascal VOC XML or COCO Text