AutoDetect-mNP is an unsupervised algorithm capable of detecting and segmenting particles from bright-field TEM images and classifying the particles based on their shapes, requiring minimum human intervention in the process. The algorithm was validated on TEM images of 4 different Au nanoparticle systems (nanorods with ~5 aspect ratio, nanorods with ~2 aspect ratio, mixture of the two nanorods, triangular nanoprisms). This dataset contains the unprocessed TEM images of the triangular prisms. For details regarding the data collection process, please refer to https://doi.org/10.1021/jacsau.0c00030.