Closed mingchuan123 closed 6 years ago
I didn't study the code in details (yet), my plan is to manipulate the code to do multiple tracking on my own custom dataset (I don't know if it's possible yet though). But I believe I can answer your question, it seems that they run _compile_results
to get information to display in the cli. IOU is calculated by using bboxes and gt, which I assume is the tracked bboxes with the groundtruth bboxes, to calculate the precision of the tracker.
Edit: They initialize the tracker with first ground truth element and then they use the network to track until the end.
Feel free to let me know if you need any more assistance, I'm analyzing the code now so I might be able to provide better help.
Thanks for your reply mike, BTW do you know how to train the model?
2017-11-02 4:36 GMT-04:00 Mike Røntved notifications@github.com:
I didn't study the code in details (yet), my plan is to manipulate the code to do multiple tracking on my own custom dataset (I don't know if it's possible yet though). But I believe I can answer your question, it seems that they run _compile_results to get information to display in the cli. IOU is calculated by using bboxes and gt, which I assume is the tracked bboxes with the groundtruth bboxes, to calculate the precision of the tracker.
Didn't find it yet but I assume they initialize the tracker with the first ground truth bounding box and then they let the siamfc do rest of the tracking.
Feel free to let me know if you need any more assistance, I'm analyzing the code now so I might be able to provide better help.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/torrvision/siamfc-tf/issues/18#issuecomment-341351923, or mute the thread https://github.com/notifications/unsubscribe-auth/AUKpyAuezGH0cdKLlhtQjKxRC32FQ84pks5syX8ggaJpZM4QIfuk .
@mingchuan123 I've decided not to look further into it, since the network is only designed for single object tracking, so I'm not able to help you further.
@mingchuan123 see #20
Hi there, I have noticed that when trying to run the model on my own images, I need to put both images and ground truth in the data folder, which seems just like ploting ground truth on the image. So how does the model track the target? Does the ground truth put in only for IoU computing? Thanks.