Short name:
OLFWMOT
Detector:
Public
Description:
we proposes leveraging wavelet decomposition to segregate target features of distinct frequency tiers. Features obtained from two-dimensional wavelet decomposition exhibit orthogonality and complementarity along the horizontal and vertical directions. Low-frequency components typically pertain to occluded targets, while high-frequency energy often emanates from targets undergoing occlusion, enabling discrimination between multi-layered objectives. Furthermore, advancing with local attention enhanced GRU learning mechanism for refactor features and further discriminating similar targets, increasing response energy in specific regions.
Reference:
Last submitted:
March 30, 2024 (1 month ago)
Published:
March 30, 2024 at 13:57:24 CET
Submissions:
3
Project page / code:
n/a
Open source:
No
Hardware:
1400MHz, 1 Core.
Runtime:
9.6 Hz
Benchmark performance:
Sequence | MOTA | IDF1 | HOTA | MT | ML | FP | FN | Rcll | Prcn | AssA | DetA | AssRe | AssPr | DetRe | DetPr | LocA | FAF | ID Sw. | Frag |
MOT20 | 57.3 | 64.1 | 48.8 | 660 (53.1) | 160 (12.9) | 100,140 | 117,943 | 77.2 | 80.0 | 48.2 | 49.8 | 54.1 | 66.8 | 60.1 | 62.2 | 77.3 | 22.4 | 2,811 (0.0) | 4,591 (0.0) |
Detailed performance:
Sequence | MOTA | IDF1 | HOTA | MT | ML | FP | FN | Rcll | Prcn | AssA | DetA | AssRe | AssPr | DetRe | DetPr | LocA | FAF | ID Sw. | Frag |
MOT20-04 | 66.7 | 72.5 | 54.5 | 427 | 44 | 50,799 | 39,492 | 85.6 | 82.2 | 53.3 | 56.0 | 59.7 | 69.6 | 67.0 | 64.4 | 77.5 | 24.4 | 967 | 1,898 |
MOT20-06 | 49.3 | 52.6 | 41.1 | 112 | 69 | 20,976 | 45,364 | 65.8 | 80.6 | 39.2 | 43.5 | 44.2 | 62.1 | 51.1 | 62.7 | 77.7 | 20.8 | 956 | 1,342 |
MOT20-07 | 55.5 | 61.9 | 46.8 | 61 | 8 | 7,849 | 6,587 | 80.1 | 77.2 | 44.7 | 49.5 | 49.7 | 64.6 | 61.0 | 58.8 | 75.0 | 13.4 | 289 | 416 |
MOT20-08 | 38.5 | 52.2 | 40.2 | 60 | 39 | 20,516 | 26,500 | 65.8 | 71.3 | 41.5 | 39.3 | 47.4 | 63.2 | 50.5 | 54.7 | 77.0 | 25.5 | 599 | 935 |
Raw data: