How to deal with no object images
Hi, I want to deal with some video frames that the object being tracked will sometimes leave the frame and gets back. How should I set the coordinates of the bounding box for no object images? I tried set them to 0,0,0,0 but then the model made a big bias on drawing a corner point even for normal images. Is this a correct way to annotate the data? Or should I have another output just to predict the possibility of whether to track or not? Thanks
The current method does not handle this, so you will need to extend the method to handle this case. Best of luck!
On Sun, Dec 3, 2017 at 11:35 PM, York [email protected] wrote:
Hi, I want to deal with some video frames that the object being tracked will sometimes leave the frame and gets back. How should I set the coordinates of the bounding box for no object images? I tried set them to 0,0,0,0 but then the model made a big bias on drawing a corner point even for normal images. Is this a correct way to annotate the data? Or should I have another output just to predict the possibility of whether to track or not? Thanks
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/davheld/GOTURN/issues/62, or mute the thread https://github.com/notifications/unsubscribe-auth/AEHoHDxC30vPObcRxoF-AFmKYmDHea7cks5s83apgaJpZM4Q0GWo .