Taco 野外废物的开放图像数据集

Taco 是一个开放的野外废物图像数据集。它包含在不同环境下拍摄的垃圾照片，从热带海滩到伦敦街道。根据分层分类法对这些图像进行手动标记和分割，以训练和评估对象检测算法。了解 TACO 的最佳方式是探索我们的数据集。 为方便起见，注释以 COCO 格式提供。 TACO 还是个婴儿，但它正在成长，您可以帮助它！我们的计划是最终开启基准挑战。敬请关注！ 数据背景： 人类一直在破坏地球，从马里亚纳海沟底部到珠穆朗玛峰。每分钟至少有 15 吨塑料垃圾泄漏到海洋中，相当于一辆垃圾车的容量。我们都看到了这种行为对野生动物的影响，海龟被塑料袋呛到，鸟儿装满瓶盖。最近的研究还在人类粪便中发现了微塑料。这些应该保留在回收链中，而不是我们的食物链中。是时候进行一场革命了。 我们相信人工智能可以发挥重要作用。想想无人机调查垃圾、机器人捡垃圾、反垃圾视频监控和 AR 来教育和帮助人类分离垃圾。这就是我们的愿景。随着深度学习的最新进展，所有这一切现在都成为可能。然而，要学习准确的垃圾检测器，深度学习需要许多带注释的图像。进入塔可。虽然还有其他一些垃圾数据集，但我们认为这些还不够。我们的目标是通过以下功能将 TACO 提升到一个新的水平： 1、对象分割。通常使用的边界框不足以完成某些任务，例如机器人抓取 2、免费许可下的图像。只要您引用我们，您就可以对 TACO 做任何您想做的事情。 3、背景注释。 TACO 涵盖了许多为了方便而标记的环境。 4、对象上下文标记。并非 TACO 中的所有物品都属于垃圾。有些物品是手持式的，甚至还不是垃圾。因此，对象是基于上下文标记的 Terms and Conditions Both the images and annotations provided by this dataset are all under free copyright licences. While the annotations are licenced under CC BY 4.0, images may be under different specific public licences since these were obtained from different sources and users. The licence along with the original URL of each individual image are referenced in the accompanying annotation file. If the licence entry is missing, then this is by default: CC BY 4.0.

Taco is an open-access field waste image dataset. It contains photographs of litter captured in diverse environments, spanning from tropical beaches to the streets of London. These images have been manually annotated and segmented using a hierarchical taxonomy for the purpose of training and evaluating object detection algorithms. The optimal way to familiarize yourself with Taco is to explore our dataset. For convenience, annotations are provided in COCO format. Taco is still in its infancy but is growing rapidly, and you can contribute to its development! Our plan is to eventually launch a benchmark challenge. Stay tuned! Data Background: Humans have been inflicting damage on the Earth, from the depths of the Mariana Trench to the summit of Mount Everest. At least 15 metric tons of plastic waste leaks into the world's oceans every minute, equivalent to the payload of a standard garbage truck. We have all witnessed the harm this causes to wildlife: sea turtles choking on plastic bags, and birds with plastic bottle caps in their digestive tracts. Recent studies have even detected microplastics in human feces. These materials should remain within recycling loops rather than entering our food chain. It is time for a revolution. We believe that artificial intelligence (AI) can play a critical role. Consider applications such as drone-based waste surveys, waste-collecting robots, anti-litter video surveillance, and AR tools to educate and assist humans in waste sorting. This is our vision. With the latest advancements in deep learning, all of these applications are now feasible. However, training accurate waste detection models with deep learning requires a large volume of annotated images. Enter Taco. While there are other existing litter datasets, we believe they are insufficient for our needs. Our goal is to elevate Taco to the next level with the following features: 1. Object segmentation. Commonly used bounding boxes are inadequate for certain tasks, such as robotic grasping. 2. Freely licensed images. You can use Taco for any purpose as long as you appropriately cite our work. 3. Background annotations. Taco covers a wide range of environments, with annotations provided for convenience across these settings. 4. Context-aware object annotation. Not all items in Taco are categorized as litter; some are handheld objects or not yet classified as waste. Accordingly, objects are annotated based on their contextual background. Terms and Conditions Both the images and annotations provided by this dataset are all under free copyright licences. While the annotations are licenced under CC BY 4.0, images may be under different specific public licences since these were obtained from different sources and users. The licence along with the original URL of each individual image are referenced in the accompanying annotation file. If the licence entry is missing, then this is by default: CC BY 4.0.

搜集汇总

背景与挑战

背景概述

Taco 野外废物的开放图像数据集是一个开放的图像数据集，包含不同环境下的垃圾照片，用于训练和评估对象检测算法。数据集以COCO格式提供注释，强调其开放许可和环境保护应用。