- 1: Extract the video frames using a frame extractor (e.g: ffmpeg)
- ffmpeg -i Video.mpg frame%d.jpg
- 2: Download and launch GAT with Sun's Java Web Start
- 3: Create a new annotation session choosing an existing ontology or creating a new on with the ontology editor on the main toolbar
- 4: Select Sequence Annotator Perspective. Add all frames using Add Images button on the main toolbar.
If the user works with videos, the frame extraction is a prerequisite step to join the tool.
- 5: Select the semantic class to be instanced in the tree
- 6: Go to the thumbnail panel and select the first and the last image of the first scene of the sequence that contains the object.
- 7: Right click on an image of the selected scene where is going to extract the object to be tracked in the rest of the images. A dialog that contains the whole image appears.
- 8: Drag the mouse to draw conveniently a rectangle around the desired object.
- 9: Right click to call the tool that performs the region-based segmentation of the defined rectangle and its representation through a hierarchical partition tree.
- 10: Choose one interactive segmentation technique and extract the desired object.
- 11: Right click to run the object tracker tool. The dialog disappears and the thumbnail panel will be showed again but noy displaying the results o the tracking tool.
- 12: Repeat from step 5 to step 10 for the rest of scenes of the sequence that contain the object.
- 13: Select the validate button
to validate the annotation
- If you have any trouble go to the
main's page "Watch it work" to see the video guide or the manual