Terminology¶
Annotate:
Manually select target objects in images and add labels to them.
Label:
The tag added to an image after annotation to identify its class.
Dataset:
The .dlkdb file containing annotated data exported by Mech-DLK.
Labeled:
The image data status of having been annotated manually.
Unlabeled:
The image data status of having not been annotated manually.
Training Set:
An image data set that has been annotated manually and is used to train the model.
Validation Set:
An image data set that has been annotated manually and is used to validate the training effect of the model.
OK Image:
A defect-free image.
NG Image:
An image with object defect.
Train:
The process of using a training set to train a deep learning model.
Validate:
The process of using a trained model to predict on the validation set and comparing the results with the validation set labels.
Accuracy:
The ratio of the number of correctly predicted samples to the total number of samples when the model predicts on a validation set.
Loss:
The degree of inconsistency between the validation set result labels from model prediction and the actual labels.
Epoch:
The number of passes of the entire training set the machine learning algorithm has completed for training.