BristolGorillas2020
收藏资源简介:
This MP4 video and JPG image dataset of a troop of 7 western lowland gorillas (Gorilla gorilla gorilla) filmed at Bristol Zoo Gardens contains around 5k+ facial bounding box and individual gorilla identity annotations. A basic YOLOv3-powered application is able to perform facial identifications at 92% mAP when utilising single frames on the still image data. Tracking-by-detection-association and identity voting across short tracklets in videos yields an improved robust performance at 97% mAP. The github repository https://github.com/obrookes/BristolGorillas2020 provides code and further info to train from scratch with this dataset. When using this dataset please cite the dataset and the paper "A Dataset and Application for Facial Recognition of Individual Gorillas in Zoo Environments" accompanying it available at https://arxiv.org/abs/2012.04689.
本数据集收录了于布里斯托动物园园区(Bristol Zoo Gardens)拍摄的7只西部低地大猩猩(Gorilla gorilla gorilla)群体的MP4视频与JPG图像数据,共包含约5000+组面部边界框(bounding box)与个体大猩猩身份标注(annotations)。一款基于YOLOv3开发的基础应用程序,在静态图像数据上使用单帧进行面部识别时,可达到92%的平均精度均值(mean Average Precision, mAP)。在视频处理中采用检测关联跟踪(Tracking-by-detection-association)以及跨短轨迹(short tracklets)身份投票的策略后,模型鲁棒性能得到显著提升,mAP达到97%。GitHub仓库(https://github.com/obrookes/BristolGorillas2020)提供了基于该数据集从零开始训练的代码与更多相关资料。使用本数据集时,请引用该数据集及其配套论文《动物园环境下的个体大猩猩面部识别数据集与应用》,论文可通过https://arxiv.org/abs/2012.04689获取。




