Change detection is one of the important tasks for video surveillance systems. A variety of learning-based approaches have been proposed, but class imbalance in training data degrades their learning efficiency. In this paper, we propose a cross entropy loss with a modulating term in cosine form to handle this class imbalance. Although the original focal loss focuses only on reducing weights for well-classified data, the proposed function is designed to preserve sufficient gradients for rare hard samples as well. This property allows a network to learn mainly from a few significant samples on which the network should focus. We validate the proposed loss through various experiments on CDNet2014 dataset, and the results show that the network trained with the proposed loss achieves better performance than other state-of-the-arts in various complex scenarios.

BACKGROUND: Several recent studies showed that next-generation sequencing (NGS)-based human leukocyte antigen (HLA) typing is a feasible and promising technique for variant calling of highly polymorphic regions. To date, however, no method with sufficient read depth has completely solved the allele phasing issue. In this study, we developed a new method (HLAscan) for HLA genotyping using NGS data. RESULTS: HLAscan performs alignment of reads to HLA sequences from the international ImMunoGeneTics project/human leukocyte antigen (IMGT/HLA) database. The distribution of aligned reads was used to calculate a score function to determine correctly phased alleles by progressively removing false-positive alleles. Comparative HLA typing tests using public datasets from the 1000 Genomes Project and the International HapMap Project demonstrated that HLAscan could perform HLA typing more accurately than previously reported NGS-based methods such as HLAreporter and PHLAT. In addition, the results of HLA-A, -B, and -DRB1 typing by HLAscan using data generated by NextGen were identical to those obtained using a Sanger sequencing-based method. We also applied HLAscan to a family dataset with various coverage depths generated on the Illumina HiSeq X-TEN platform. HLAscan identified allele types of HLA-A, -B, -C, -DQB1, and -DRB1 with 100% accuracy for sequences at ≥ 90× depth, and the overall accuracy was 96.9%. CONCLUSIONS: HLAscan, an alignment-based program that takes read distribution into account to determine true allele types, outperformed previously developed HLA typing tools. Therefore, HLAscan can be reliably applied for determination of HLA type across the whole-genome, exome, and target sequences.

This paper proposes a novel path prediction algorithm for progressing one step further than the existing works focusing on single target path prediction. In this paper, we consider moving dynamics of co-occurring objects for path prediction in a scene that includes crowded moving objects. To solve this problem, we first suggest a two-layered probabilistic model to find major movement patterns and their cooccurrence tendency. By utilizing the unsupervised learning results from the model, we present an algorithm to find the future location of any target object. Through extensive qualitative/quantitative experiments, we show that our algorithm can find a plausible future path in complex scenes with a large number of moving objects.