When you use the auto label function on a data set, only frames and images are processed. Videos are ignored. However, you can run the auto label function on an individual video.
About this task
Any frames that were previously captured by using auto capture and were not manually labeled are deleted before auto labeling. This deletion helps avoid labeling duplicate frames. Manually captured frames are not deleted.
Complete the following steps to run the auto label function on a video.
Procedure
-
Open the data set that contains the video.
-
Select the video and click Label objects.
-
Click Auto label, choose the appropriate settings, then click Auto label
Frames are captured at the specified interval and then the specified trained model is used to process the frames. When an object is identified with the specified confidence threshold, it is labeled. By default, the automatically added labels are light red.
-
Review the automatically added labels on the Data Set page. You can move or resize the labels that were automatically generated. You can also save or reject individual labels, or you can reject them all by selecting Clear all. Saving or manipulating a label converts it to a manually added label. Rejecting a label deletes it. If you run Auto label again, any images or frames that now have manually added labels are skipped.
You can use the confidence filter in the filter bar to review labels by the confidence level that is assigned by the model that was used to auto label the data set. For example, you can filter low confidence labels, which are likely wrong, and easily reject them, or filter on high confidence labels to quickly accept labels that are probably accurate.
Note: Auto labels that were created before IBM® PowerAI Vision 1.1.5 do not have a confidence value. These auto labels are treated as 0% confidence. Therefore, when using the confidence filter, they are displayed only if the minimum confidence is set to 0.