Dimension types for different CV tasks

  CV Tasks , Image annotation is helpful for computers to better understand images . The computer will, based on known label information , Learn the similar rules applicable to new data identification from the data type .

CV Annotation is divided into the following ways ：

• Border box marking
• Polygon annotation
• Key point marking
• Line marking
• Box dimension （3D）
• Semantic segmentation

Border box marking

Bounding box is the most common type of image annotation . As it literally means , The annotator needs to draw a box around the target object according to the specific requirements . You can use the bounding box to train the target detection model .

Polygon annotation

Polygonal mask （mask） It is mainly used to mark targets with irregular shapes . The annotator must mark the boundary of the object in the image with high precision , So as to clearly understand the shape and size of the target . It is different from the dimensioning method of dimension box , You can frame unnecessary areas around the target, which may affect the training of the model in some tasks , Polygon annotation can obtain more accurate positioning results in the task because of its high dimensioning accuracy .

Key point marking

Landmark The marking is mainly applicable to Visual tasks to detect shape changes and small objects , It helps to better understand the motion changes of each point in the target object . Key point annotation can help realize gesture and face recognition , It can also be used to detect body parts and accurately estimate their posture .

Line marking

Line dimensioning is done by drawing lane line annotations to apply to Training vehicle perception model task for lane detection . Unlike bounding boxes , It avoids a lot of white space and extra noise .

Box dimension

3D Box annotation is a visual task used to calculate the depth of the target object , Such as vehicle , Buildings and even people , To obtain its total volume . It is mainly used in the field of construction and autonomous vehicle systems .

Semantic segmentation

In semantic segmentation or pixel level annotation , We combine pixels with similar properties . It applies to Visual task of detecting and locating specific targets at pixel level . And used to detect specific target objects （ Or areas of interest ） Different polygon segmentation , Semantic segmentation provides a complete understanding of each pixel of the scene in the image .