A training data set is a collection of data used to train a machine learning model to detect objects, classify images, notice anomalies, or perform other operations.
A training clip refers to captured video data whose image frames are added to a data set for labeling.
In the "ML" tab, select your model.
In your model's "Training Data Sets" tab, click "Create Data Set"
Enter a name for your data set and optional description.
If desired, check "Send images to HyperLabel automatically as they are added"
Click "Save & Continue to Add Training Data"
Select the camera whose video you want to build training data from
Select the video segment you want to label
Click "Save & Continue to Create Training Clips"
To select a specific portion of this segment for labeling, click "Add Clip"
Set a Start Time and End Time using the slider, or manually enter the interval.
When you have selected your clip, click "Create Training Data"
This will split the clip into frames depending on your camera's frame rate (fps).
Note: Data sets will remain in "Pending" state until the system has processed the clip and cut them into individual images. Depending on the size of the clip, this could take several minutes. Once complete, they will be marked "Processed" after the page is refreshed.
You can also create training clips under the Camera Recordings table of the Camera information page.
In Devices, go to the "Cameras" tab.
Navigate to your camera.
Scroll down to the Camera Recordings section. Here you can filter your recordings by a date range.
Find your desired video segment and click the "Create Clips" link under the "Actions" column.
You'll be taken to the Create Training Clips view described above where you can select a portion of your training clip for labeling.