Evaluation
Trackers

We evaluate 10 state-of-the-art tracking algorithms on BSPT.

The Characteristics of Compared Trackers
No. Trackers Feature Scale estimation Offline training Online Learning GPU Publication Year Resource Link Implementation
1 ATOM CNN Yes Yes Yes Yes CVPR 19 link Python+torch
2 DiMP CNN Yes Yes Yes Yes ICCV 19 link Python+torch
3 ECO CNN, HOG, CN Yes No Yes Yes CVPR 17 link MATLAB+matconvnet
4 SiamFC CNN Yes Yes No Yes ECCV 16 link MATLAB+matconvnet
5 SiamMask CNN Yes Yes No Yes CVPR 19 link Python+torch
6 SiamRPN CNN Yes Yes No Yes CVPR 18 link Python+torch
7 SiamRPN++ CNN Yes Yes No Yes CVPR 19 link Python+torch
8 UPDT CNN, HOG, CN Yes Yes Yes Yes ECCV 18 link MATLAB+matconvnet
9 MDNet CNN Yes Yes Yes Yes CVPR 16 link MATLAB+matconvnet
10 KYS CNN Yes Yes Yes Yes ECCV 20 link Python+torch
Overall performance
Attribute-base performance
Precision Evaluation on different attributes
　 STO (73) DTO (105) POC (111) FOC (69) MB (76) ROT (68) DEF (44) OV (30) overall
KYS 0.582 0.63 0.631 0.594 0.641 0.645 0.548 0.755 0.65
UPDT 0.581 0.639 0.636 0.595 0.627 0.63 0.582 0.661 0.65
SiamRPN++ 0.509 0.576 0.576 0.521 0.579 0.577 0.506 0.727 0.6
ECO 0.58 0.625 0.631 0.602 0.622 0.633 0.587 0.617 0.645
MDNet 0.547 0.62 0.628 0.548 0.596 0.597 0.578 0.551 0.613
SiamFC 0.534 0.584 0.589 0.553 0.591 0.589 0.525 0.761 0.61
DiMP50 0.476 0.501 0.509 0.501 0.541 0.502 0.471 0.725 0.536
SiamMask 0.521 0.553 0.553 0.556 0.546 0.568 0.519 0.615 0.569
ATOM 0.464 0.486 0.492 0.465 0.507 0.498 0.447 0.71 0.522
DiMP18 0.426 0.454 0.466 0.439 0.482 0.443 0.416 0.698 0.491
Success Evaluation on different attributes
　 STO (73) DTO (105) POC (111) FOC (69) MB (76) ROT (68) DEF (44) OV (30) overall
KYS 0.365 0.393 0.397 0.352 0.412 0.4 0.342 0.517 0.414
UPDT 0.336 0.368 0.368 0.328 0.371 0.362 0.323 0.394 0.378
SiamRPN++ 0.319 0.352 0.355 0.303 0.37 0.344 0.297 0.484 0.375
ECO 0.327 0.355 0.361 0.327 0.361 0.358 0.326 0.363 0.37
MDNet 0.305 0.354 0.361 0.3 0.343 0.339 0.32 0.308 0.351
SiamFC 0.303 0.332 0.337 0.307 0.34 0.333 0.287 0.451 0.35
DiMP50 0.297 0.316 0.324 0.297 0.35 0.317 0.292 0.501 0.347
SiamMask 0.316 0.326 0.329 0.316 0.336 0.331 0.3 0.386 0.342
ATOM 0.29 0.295 0.302 0.272 0.32 0.305 0.271 0.473 0.326
DiMP18 0.271 0.288 0.299 0.261 0.316 0.283 0.264 0.479 0.319
Download

The raw results can be downloaded in:

- Download

 

Laboratory of Intelligent Media Computing and Network Security，HUST. ©  2021. All rights reserved.