Video data and algorithms have been driving advances in multi-object tracking (MOT). While existing MOT datasets focus on occlusion and appearance similarity, complex motion patterns are widespread yet overlooked. To address this issue, we introduce a new dataset called BEE23 to highlight complex motions. Identity association algorithms have long been the focus of MOT research. Existing trackers can be categorized into two association paradigms: single-feature paradigm (based on either motion or appearance feature) and serial paradigm (one feature serves as secondary while the other is primary). However, these paradigms are incapable of fully utilizing different features. In this paper, we propose a parallel paradigm and present the Two rOund Parallel matchIng meChanism (TOPIC) to implement it. The TOPIC leverages both motion and appearance features and can adaptively select the preferable one as the assignment metric based on motion level. Moreover, we provide an Attention-based Appearance Reconstruct Module (AARM) to reconstruct appearance feature embeddings, thus enhancing the representation of appearance features. Comprehensive experiments show that our approach achieves state-of-the-art performance on four public datasets and BEE23. Notably, our proposed parallel paradigm surpasses the performance of existing association paradigms by a large margin, e.g., reducing false negatives by 12% to 51% compared to the single-feature association paradigm. The introduced dataset and association paradigm in this work offer a fresh perspective for advancing the MOT field. The source code and dataset are available at https://github.com/holmescao/TOPICTrack.

In geophysics, crowdsourcing is an emerging non-traditional environmental monitoring approach that encourages contributions of data from individual citizens. Because of their reliance on undertrained citizens and imprecise low-cost sensors, crowdsourced data applications suffer from different types of noises that can deteriorate the overall monitoring accuracy. In this study, we propose a machine learning approach for automatic Crowdsourced data Quality Control (CSQC) by detecting and removing noisy data points in spatially and temporally discrete crowdsourced observations. We design a set of features from the original and interpolated rainfall data, and apply them to train and test the CSQC models based on both supervised and non-supervised machine learning algorithms. Performances of the CSQC models under various scenarios assuming no further retraining are also tested (hereafter referred to as transferability). The results based on synthetic but realistic data show that the CSQC model can significantly reduce the overall rainfall estimation error. Under the stationary assumption, CSQC models based on both supervised and unsupervised algorithms can have decent performances in noisy data identification and overall rainfall estimation error reduction; however, if the model is transferred to other cities with different rainfall structure or noise composition (without retraining), the supervised Multi-Layer Perceptrons (MLPs) turns out to be the best performing one.