Label the Training Data

Create Label(s)

Please create label(s) based on project needs. For instance, if the project need is distinguishing the front and back faces of a workpiece, please create the labels “front” and “back”.

Attention

Label names should be relevant to the objects and easily recognizable. Please do not use meaningless names like a, b, tmp, etc. Label names should only include letters or numbers.

Determine Method of Labeling

1. If the task is to distinguish different parts of a single object, please label the prominent feature for each part with boxes, as shown below.

   

   Figure 1. Labeling the prominent features for different parts of a single object

2. If the task is to distinguish objects of different classes, please envelop the entire object with a box, as shown below.

   

   Figure 2. Labeling different objects with boxes

3. If the input images for the actual application have the background removed, please label the entire contour of each object in the training dataset, as shown below.

   

   Figure 3. Labeling the entire contours of objects

Attention

Please ensure the quality of labeling. Any incorrect labels will adversely affect model performance. For instance, if in ten images of the workpiece’s front face, one is labeled as “back”, the classification performance will be severely affected.