July 5, 2019
How to pick the right type of annotation for Computer Vision Machine Learning Models
Annotation in Machine Learning is the process of labeling the data, which could be in the form of text, video, images or audio. Image annotation helps to make images readable for computer vision, computers use the annotated data to learn to recognize similar patterns when presented with new data.
Bounding Box Annotation
Bounding boxes are the most common type of image annotation. As it sounds like, labeler have to draw a box around the objects of interest based on specific requirements. Object localization and object detection models can be trained using bounding boxes.
The Polygonal masks are mainly used to annotate objects with irregular shapes. Labelers must generate boundaries of objects in a frame with high precision, this gives a clear idea about the shape and size of the object. Unlike boxes, which can capture unnecessary objects around the target, which leads to confusing the training of your computer vision models, polygons are more accurate when it comes to localization.
Landmark or Key-point Annotation
The Landmark annotation is used to detect shape variations and small objects, it helps to better understand each point motion in the targeted object. Key points can help in gesture and facial recognition and is also used to detect the human body parts and estimate their poses with right accuracy.
The Line Annotation is used to draw lanes to train vehicle perception models for lane detection. Unlike the bounding box, it avoids many blank spaces and additional noises.
The 3D cuboids are used to calculate the depth of the targeted objects as vehicles, buildings or even humans to get their overall volume. It is mainly used in the field of construction and autonomous vehicle systems.
In Semantic Segmentation or Pixel-level annotation we group together the pixels that have similar attributes.It is used for detection and localization of specific objects in pixel-level. Unlike polygonal segmentation used to detect specific objects of interest, full semantic segmentation provides a complete understanding of every pixel of the scene in the image.
Dealing with large datasets and need to scale up your annotation efforts? Feel free to try LinkedAI Platform or get in touch with us at email@example.com— our annotators would love to contribute to your project.