Aggregation Signature for Small Object Tracking

Abstract

Small object tracking becomes an increasingly important task, which however has been largely unexplored in computer vision. The great challenges stem from the facts that: 1) small objects show extreme vague and variable appearances, and 2) they tend to be lost easier as compared to normal-sized ones due to the shaking of lens. In this paper, we propose a novel aggregation signature suitable for small object tracking, especially aiming for the challenge of sudden and large drift. We make three-fold contributions in this work. First, technically, we propose a new descriptor, named aggregation signature, based on saliency, able to represent highly distinctive features for small objects. Second, theoretically, we prove that the proposed signature matches the foreground object more accurately with a high probability. Third, experimentally, the aggregation signature achieves a high performance on multiple datasets, outperforming the state-of-the-art methods by large margins. Moreover, we contribute with two newly collected benchmark datasets, i.e., small90 and small112, for visually small object tracking. The datasets will be available in https://github.com/bczhangbczhang/.

A Comprehensive Review for Typical Applications Based Upon Unmanned Aerial Vehicle Platform

“…Website Source Okutama-Action [24] http://okutama-action.org VisDrone2018-2022 [25] https://github.com/VisDrone/VisDrone-Dataset MOR-UAV [26] https://visionintelligence.github.io/Datasets.html CARPK [27] https://paperswithcode.com/dataset/carpk AU-AIR [28] https://bozcani.github.io/auairdataset DroneVehicle [30] https://github.com/VisDrone/DroneVehicle UVSD [29] https://github.com/liuchunsense/UVSD UAVDT [34] https://sites.google.com/view/grli-uavdt/ BIRDSAI [31] https://sites.google.com/view/elizabethbondi/dataset Stanford Drone Dataset [35] https://cvgl.stanford.edu/projects/uav data/ HighD [36] https://www.highd-dataset.com DTB70 [37] https://github.com/flyers/drone-tracking UAV123/20L [38] https://cemse.kaust.edu.sa/ivul/uav123 Anti-UAV [39] https://github.com/ucas-vg/Anti-UAV Small90/112 [40] https://github.com/bczhangbczhang/smallobject UAVDark135 [41] https://vision4robotics.github.io/project/uavdark135/ DarkTrack2021 [42] https://darktrack2021.netlify.app UAVTrack112 [43], [44] https://github.com/vision4robotics/SiamAPN AVSD [45] https://github.com/wyfeng1020/AVSD UAVid [46] https://uavid.nl AeroScapes [47] https://github.com/ishann/aeroscapes ManipalUAVid [48] https://github.com/uverma/ManipalUAVid (49,712) and arbitrary quadrilateral bounding boxes (47,519 small vehicles and 2193 large vehicles). The resolution of these data includes 4000×3000, 5472×3648, and 4056×3040.…”

“…The videos cover a variety of backgrounds (e.g., tree, cloud, building), two light modes (visible and infrared) and two lighting conditions (night and day) at 25 FPS. Small90/112: Small90 [40] comprises 90 small-sized object sequences with about 39,380 frames, in which additional challenges (e.g., low resolution and target drifting) are encompassed. Based on Small90, Small112 [35] adds another 20 more challenging sequences.…”

COMET: Context-Aware IoU-Guided Network for Small Object Tracking

“…Extensive experimental analyses are performed to compare the proposed tracker with state-of-the-art methods on recent and related small object tracking benchmarks, namely, UAVDT [13], VisDrone-2019 [14], and Small-90 [39].…”

“…In this section, the proposed method is evaluated on three state-of-the-art benchmarks for small object tracking from aerial view: VisDrone-2019 (35 videos) [14], UAVDT (50 videos) [13], and Small-90 (90 videos) [39]. Video sequences in the VisDrone-2019, UAVDT, and Small-90 benchmarks are annotated by twelve, eight, and eleven attributes, respectively.…”

Do We Really Need Frame-by-Frame Annotation Datasets for Object Tracking?

“…This benchmark defined several new evaluation metrics such as EAO, Accuracy, and Robustness, broadly adopted nowadays in the tracking community. Moreover, there are datasets made for some specific tasks such as small object [25] and UAV tracking [28]. However, the scale of these datasets is usually small, limiting the further progress of high-complexity tracking models.…”