This is a great repository and I've been working with an offshoot from mayuri0192 which uses LSTM to track an object using location prediction rather than feature identification. The videos I'm trying to track in work much better if the object is manually selected and we have some decent tracks using openCV; would it be possible to replace the yolo_out file with a variation of location outputs from the openCV tracker? How is the yolo_out file created to begin with?
This is a great repository and I've been working with an offshoot from mayuri0192 which uses LSTM to track an object using location prediction rather than feature identification. The videos I'm trying to track in work much better if the object is manually selected and we have some decent tracks using openCV; would it be possible to replace the yolo_out file with a variation of location outputs from the openCV tracker? How is the yolo_out file created to begin with?
Thanks!