Open gongbudaizhe opened 4 years ago
Hi, Due to our code is based on atom, you need to implement ATOM performance on your computer first.
I have reproduced the performance of ATOM which is 0.401. However, using the same atom model with the SPS modification, the EAO is 0.389.
Moreover, can you repeat your experiment on VOT2018 15 times and upload the results?
Hi, You can get our results from https://drive.google.com/open?id=1IPpOGYU6r5Dmz20PcrfOBtMCJ60C6q3Y .
Hi,
I evaluated the SPSTracker 15 times on VOT2018 just like ATOM. However, the EAO is 0.389 which is far lower than the reported 0.434 in the paper. Even worse than the ATOM tracker which has EAO 0.401.
Moreover, I notice the uploaded VOT raw results contains only 1 run. This is not standard VOT evaluation which requires 15 repetitions for stochastic trackers.