When they should be used: Opted for situations when the object doesn’t fit well in a rectangle due to its irregular shape (such as bodies of waters, land areas, etc), size or the direction of the image or developers need a more precise annotation
Pros:
Extremely handful for projects with complex irregularly shaped objects
Create pixel-perfect annotations
Cons:
Taking a considerably longer time to annotate than bounding box tools
Too much point can create lagging when computing
Using the rectangle to enclose the target object in the image or video. The box should be as near to every edge of the item as possible.
When they should be used: When the item’s shape isn’t important or when occlusion isn’t a problem.
Pros:
Simple and quick to draw, easily save time for enormous data projects
One of the most common and cheapest annotation tools to compute
Cons:
Only restricted for rectangles and squares
Cannot precisely only the objects’ pixels but rather includes backgrounds, other entities’ parts, etc
Bad for composite and occluded objects.
Consisting of a series of points (key points), connected by lines. TagOn also provides pre-developed keypoint frame for common projects such as skeleton, facial expressions, etc
When they should be used: Useful for movement tracking and prediction, keeping track of differences between objects that always have the same structure (e.g. human figures and facial features). Therefore they are commonly used in face recognition, recognizing body parts, postures, and facial emotions.
Pros:
Quick and simple when using our pre-developed keypoint frames: annotators only need to align the point to objects
Ensuring that no node in the sequence is overlooked.
Cons:
Only useful for objects with the same regular defined structures
Might cause confusion for annotators when the nodes of the objects are not visible or hidden
3d bounding boxes encapsulate the objects and display them with depth, allowing computer vision algorithms to interpret volume and orientation. For annotators, drawing cuboids means placing 2 rectangles (front and back) and 2 shapes will automatically connect to create a 3D cube.
When it should be used: Used when depth perception is critical and can be applied in a vast range of industries: 3D structure detection in Geospatial, Lidar point cloud data for vehicle tracking, analyzing CTs and MRIs in medical, etc
Pros:
Providing extra information about the depth of the object rather than just length and width in 2D bounding boxes.
Cons:
Time-consuming, require more skills to draw and more computing capacity to process a large amount of data
Dividing audio recordings into small pieces to easily annotate data at the highest accuracy
Easily highlight document components (keywords, phrases, or sentences) with including tags to add meaning to text data
Capitalizing pre-trained tools to annotate your complex images and videos at speed. Ensuring precision with highly-detailed auto edge-segmentation to pixel levels.
Calculate distances in the image in pixels to define objects features and tailor to requestors’ demands
Create a compound shape by combining all the selected shapes into a single larger one
Allowing you to use any top/back object to create a cutout from the one underneath/above. This option is used to subtract areas of an illustration by making adjustments to the stacking order
