AmSo Advanced Dataset Labeler

Advanced object annotation

Manual (top) and AI (bottom) tools

Labeling Projects

Labeling Project Settings

Quick Browsing

AI Quality Check

Conversion Projects

Conversion Project Settings

Merge Projects

Merge Project Settings

Overview

AmSo Advanced Dataset Labeler is a tool for creating and maintaining, converting, and merging annotated images and video frames to train Object Detection models. Its design and editing features are for rapid annotations and it is backed by a powerful AI engine and related tools to assist with that task.

Label Images

Object labeling (annotating) is at the heart of AmSo Advanced Dataset Labeler and given it is something that may be performed in significant quantities, the tool is intentionally designed to make it easy and painless to you. It allows for viewing, assigning, or replacing labels, filtering images, and creation of images from video files (while keeping track of the connection to the video) all within one single screen. And it does it without use of popup or other design elements that would break your flow.

You can narrow down the files you are working on:

Training Set — View images by their training set.

All/Labeled/Unlabeled — Filter images by whether they are labelled yet or not.

Labels/Files — Filter images by a specific label.

Video — Browse videos frame by frame and label them as images.

Label new items manually and track your changes:

Drawing Tools — Custom drawing tools for bounding boxes, oriented bounding boxes, and two ways of selecting polygons.

New Star — Highlight new or modified additions. (either by you or by AI)

Reduce on-screen clutter by choosing how labels are shown and use HDR to see bounding objects more easily.

See labels as text or with images and choose favorite/common items.

List/Icons — You can choose to see labels as text or images.

Grab Icon — Grab representative images for each label.

Add label — You can add and use new labels as needed.

Control when and how changes are saved. (While editing each image it is also possible to revert it or undo changes)

Auto-Save — Traditional Save or use auto-Save to avoid popup confirmations.

AI Assisted Image Labeling

AmSo Advanced Dataset Labeler allows the AI models you have built with your data to assist with labeling of more data. AI can be used to detect objects to annotate, label them, or optimize or quality check work already performed.

AI Pinpoint — AI powered replacement for the drawing tools, click and AI will detect, outline, and label the object at that point.

AI Labeling — AI label all objects, selected ones, or only when adding new ones.

AI Detection — Detect all undetected objects on the image.

AI Optimization — AI optimize current detection outlines.

Convert Datasets

AmSo Advanced Dataset Labeler can read datasets and convert them to a different dataset format. It can also perform a number of filtering or transformation to images, labels, and annotations in the process.

Merge Datasets

AmSo Advanced Dataset Labeler can read two or more datasets and merge them together, performing filtering or transformation on images, labels, and annotations. And handling conflicts that could arise.

Security

Unlike downloaded apps which have unlimited access to your hardware, AmSo Advanced Dataset Labeler is installed and works within the MacOS sandbox giving you full permissions control. It also minimizes use of third party libraries, significantly reducing chance of misbehavior by a buggy external library.

Notes

Convert and Merge generally add to or overwrite the destination folder. While they request an empty destination directory to start with, they do not enforce it thereafter. For the final conversion or merge you should start with an empty folder again.
AmSo Advanced Dataset Labeler requires macOS Ventura or newer.
Please see support page for contact and support information.