Natural Sprites

该 csv 由 2019 年 YouTube-VIS 挑战赛中保持纵横比 (64 x 128) 的缩小二进制掩码的三个视图（左、中、右）的 (x-position, y-position, area) 元组组成，可以是可在 https://competitions.codalab.org/competitions/20127#participate-get-data 找到。从此 csv 中提取对会在给定统计数据中产生 234,652 个转换。这些统计数据可用于通过自然转换来增强地面真实因子分布，我们在 spriteworld 中对此进行了演示。有关详细信息，请参阅我们的论文，该论文可在 https://openreview.net/forum?id=EbIDjBynYJ8 找到。

This CSV consists of (x-position, y-position, area) tuples from three views (left, middle, right) of downsampled binary masks with preserved aspect ratio (64 × 128) from the 2019 YouTube-VIS Challenge, and is available at https://competitions.codalab.org/competitions/20127#participate-get-data. Extracting pairs from this CSV yields 234,652 transformations under the given statistics. These statistics can be used to augment the ground-truth factor distribution via natural transformations, which we demonstrate in the spriteworld environment. For more details, please refer to our paper, which is accessible at https://openreview.net/forum?id=EbIDjBynYJ8.

搜集汇总

背景与挑战

背景概述

Natural Sprites数据集是一个CSV文件，包含从YouTube-VIS挑战赛的二进制掩码中提取的(x-position, y-position, area)元组，涉及左、中、右三个视图。这些数据用于通过自然转换增强地面真实因子分布，相关研究基于图宾根大学2020年发布的论文。