WIDER FACE
收藏资源简介:
WIDER FACE数据集是目前最大的面部检测数据集,由香港中文大学信息工程系创建,包含32,203张图像和393,703个标注面部边界框。该数据集因其面部在尺度、姿态和遮挡方面的高度变异性而极具挑战性。数据集根据60个事件类别组织,随机选取40%/10%/50%的数据作为训练、验证和测试集。WIDER FACE数据集不仅用于面部检测算法的训练和评估,还旨在解决面部检测中的小尺度、遮挡和极端姿态等实际问题,适用于多种真实世界应用场景,如监控摄像头捕捉的面部检测。
WIDER FACE Dataset is currently the largest face detection dataset, created by the Department of Information Engineering of The Chinese University of Hong Kong. It contains 32,203 images and 393,703 annotated face bounding boxes. This dataset is highly challenging due to the extreme variability of faces in terms of scale, pose and occlusion. The dataset is organized into 60 event categories, with 40%, 10% and 50% of the data randomly selected as the training, validation and test sets respectively. The WIDER FACE Dataset not only serves for the training and evaluation of face detection algorithms, but also aims to address practical issues in face detection including small-scale faces, occlusion and extreme poses, and is applicable to a variety of real-world application scenarios such as face detection captured by surveillance cameras.

- WIDER FACE数据集首次发表,由香港中文大学多媒体实验室发布,旨在解决人脸检测中的挑战性问题。
- WIDER FACE数据集首次应用于CVPR(计算机视觉与模式识别会议),成为人脸检测领域的重要基准。
- WIDER FACE数据集的扩展版本发布,增加了更多的标注信息和图像,进一步提升了数据集的多样性和复杂性。
- WIDER FACE数据集被广泛应用于各种人脸检测算法的研究和评估,成为该领域的标准数据集之一。
- WIDER FACE数据集的性能评估方法进一步完善,引入了新的评价指标,以更全面地衡量人脸检测算法的性能。
- WIDER FACE数据集的社区贡献显著增加,多个研究团队基于该数据集提出了新的算法和模型,推动了人脸检测技术的发展。
- 1WIDER FACE: A Face Detection BenchmarkUniversity of Chinese Academy of Sciences · 2016年
- 2Face Detection with the Faster R-CNNMicrosoft Research · 2017年
- 3S3FD: Single Shot Scale-invariant Face DetectorTsinghua University · 2017年
- 4FaceBoxes: A CPU Real-time Face Detector with High AccuracyTsinghua University · 2017年
- 5RetinaFace: Single-stage Dense Face Localisation in the WildImperial College London · 2020年



