Run the detection process to determine where the objects are located in a given image.
In this case, an additional file with annotations is required for the model's training set. The annotation stores information about the labels describing the depicted objects and possible additional information, depending on the selected standard.
To view the annotations, go to the
OBJECT DETECTION tab of the
In this section, you should see a bounding box around all the objects of interest in each image.
Keep in mind: If there is an object in the photo and you do not select it during training, the model will misread the contents of the photo which will affect its overall precision (accuracy).
Not every uploaded image has been annotated, so select
Unannotated images to find all the images without object information. Click on the first one.
To add a new annotation to a photo, simply hover the cursor over the photo and create a square that covers the entire object. To make this task easier, two guide lines appear from the cursor position.
Once your new bounding box is drawn, select
FLASHLIGHT. To see the categories currently in use, expand another panel from the right hand menu.
Done. If you make a mistake, you can always check the bounding box and change the categories. Use the arrows at the top of the page or the keyboard shortcuts
D to move to the next photo.
In the next photo, select the USB adapter and mark it as
USB_ADAPTER. After marking all the photos, return to the dataset.
All of the files now have appropriate annotations.