Synthetic Cell Dataset

该数据集是一个按需生成的合成数据集，包含了具有不同特征的细胞图像以及相应的真实热力图，用于评估深度神经网络中的可解释性方法。该数据集包含10类细胞，这些细胞在形状、颜色和背景上存在变化。每个真实热力图旨在突出显示区分所有类别的特征，并根据需要可进行调整。规模上，数据集总共包含6400个训练样本，1600个验证样本以及1600个评估样本。任务是对使用生成的热力图在深度神经网络中评估可解释性方法。

This is an on-demand generated synthetic dataset containing cell images with diverse characteristics and their corresponding ground-truth heatmaps, which is specifically designed for evaluating interpretability methods in deep neural networks. The 10 cell classes included in this dataset vary in shape, color and background. Each ground-truth heatmap is intended to highlight the discriminative features that distinguish all cell classes, and can be adjusted as needed. Regarding the dataset scale, it contains a total of 6400 training samples, 1600 validation samples, and 1600 evaluation samples. The core task of this dataset is to evaluate interpretability methods in deep neural networks by leveraging the generated heatmaps.