地球资源数据云——数据资源详情
篮球计算机视觉
该数据集《Basket Ball Computer Vision》主要用于监督学习任务，数据形态以图像为主，应用场景偏向安全检测。 题目说明：Object recognition with basketball image 任务类型：图像监督学习。 建议流程：先检查类别分布与脏样本，再用迁移学习（如 ResNet/EfficientNet）建立基线。 评估建议：使用分层切分或交叉验证，优先关注 F1、Recall、AUC 等分类指标。 可用文件：未检测到标准 CSV，可优先查看目录中的索引或说明文件。 Context I'm interested in object recognition, and trying to learn how to build CNN models able to recognize specific objects. Content So I created this data set by taking pictures around my flat with a basketball in each image. The data is made up of several pairs of image and annotation files (.annote). The annotation file subdivides the image into 200x200 pixel windows and says whether basketball is present in it or not. Acknowledgements The initial analysis has used the mnist deep CNN model given in the tensor flow example adapted for multiple channels and image resolution
摘要概览
该数据集《Basket Ball Computer Vision》主要用于监督学习任务，数据形态以图像为主，应用场景偏向安全检测。 题目说明：Object recognition with basketball image
任务类型：图像监督学习。
建议流程：先检查类别分布与脏样本，再用迁移学习（如 ResNet/EfficientNet）建立基线。
评估建议：使用分层切分或交叉验证，优先关注 F1、Recall、AUC 等分类指标。
可用文件：未检测到标准 CSV，可优先查看目录中的索引或说明文件。
Context
I'm interested in object recognition, and trying to learn how to build CNN models able to recognize specific objects.
Content
So I created this data set by taking pictures around my flat with a basketball in each image. The data is made up of several pairs of image and annotation files (.annote). The annotation file subdivides the image into 200x200 pixel windows and says whether basketball is present in it or not.
Acknowledgements The initial analysis has used the mnist deep CNN model given in the tensor flow example adapted for multiple channels and image resolution