We propose a self-supervised method for incrementally refining visual descriptors to improve performance in the task of object-level visual data association. Our method optimizes deep descriptor generators online, by fine-tuning a widely available network pre-trained for image classification. We show that earlier layers in the network outperform later-stage layers for the data association task while also allowing for a 94% reduction in the number of parameters, enabling the online optimization. We show that choosing positive examples separated by large temporal distances and negative examples close in the descriptor space improves the quality of the learned descriptors for the multi-object tracking task. Finally, we demonstrate a MOTA score of 21.25% on the 2D-MOT-2015 dataset using visual information alone, outperforming methods that incorporate motion information.

中文翻译：

---

通过自标记三元组在线描述符的增强，以实现可视数据关联

我们提出了一种自我监督的方法，用于逐步细化视觉描述符，以提高对象级视觉数据关联任务的性能。我们的方法通过微调针对图像分类进行预训练的广泛可用网络，在线优化深度描述符生成器。我们显示，在数据关联任务中，网络中的较早层胜过较后层，同时还使参数数量减少了94％，从而实现了在线优化。我们表明，选择在较大的时间距离内分隔开的正样本和在描述符空间中闭合的负样本可以提高多对象跟踪任务的学习描述符质量。最后，我们仅使用视觉信息就在2D-MOT-2015数据集上证明了MOTA分数为21.25％，