Multi-object tracking (MOT) aims at estimating bounding boxes and identities of objects in videos. Most methods obtain identities by associating detection boxes whose scores are higher than a threshold. The objects with low detection scores, e.g. occluded objects, are simply thrown away, which brings non-negligible true object missing and fragmented trajectories. To solve this problem, we present a simple, effective and generic association method, tracking by associating every detection box instead of only the high score ones. For the low score detection boxes, we utilize their similarities with tracklets to recover true objects and filter out the background detections. When applied to 9 different state-of-the-art trackers, our method achieves consistent improvement on IDF1 score ranging from 1 to 10 points. To put forwards the state-of-theart performance of MOT, we design a simple and strong tracker, named ByteTrack. For the first time, we achieve 80.3 MOTA, 77.3 IDF1 and 63.1 HOTA on the test set of MOT17 with 30 FPS running speed on a single V100 GPU. The source code, pre-trained models with deploy versions and tutorials of applying to other trackers are released at https://github.com/ifzhang/ByteTrack.
Several layer-structured vanadates of two-dimensional
(2D) nanosheet
morphologies have been investigated recently for flexible quasi-solid-state
aqueous zinc-ion batteries (ZIBs), where one of the challenging issues
is the poor electronic conductivity and mechanical stability especially
in the cross-2D nanosheet direction, leading to insufficient rate
capability and mechanical stability and shortened cycle life. Herein,
we have devised a strategy of using one-dimensional (1D) carbon nanotubes
(CNTs) to stitch zinc pyrovanadate (Zn3(OH)2V2O7·2H2O, CNT-stitched ZVO)
2D nanosheets that are directly grown on oxidized CNT fiber (CNT-stitched
ZVO NSs@OCNT fiber). With the CNT-stitched 2D nanosheet structure,
the open frameworks of ZVO provide required spacing for reversible
Zn2+ (de)intercalation, and the stitching CNTs offer the
desperately needed electronic conductivity and mechanical robustness
across the ZVO 2D nanosheets. As a result, the fiber-shaped quasi-solid-state
ZIB, assembled using the CNT-stitched ZVO NSs@OCNT as the cathode
and Zn NSs@CNT fiber (electrodeposited zinc nanosheets on CNT fiber)
as the anode, demonstrates an ultrahigh rate capability (69.7% retention
after a 100-fold increase in current density), an impressively stack
volumetric energy density of 71.6 mWh cm–3, together
with a long-term stability (88.6% retention after 2000 cycles). The
present work proves the proof-of-concept of developing 2D nanosheets
purposely stitched together by 1D conducting nanotubes/nanowires as
a class of advanced cathodes for quasi-solid-state ZIBs in future
portable electronics.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.