The training data for the detection part

[jhb86253817]

Hi, could you give more details on the training images of the detection part? It seems kaggle.json is the bounding box annotation, and the images are from Real-Life-Violence-Dataset, but the image names are not consistent. By the way, did you annotate the bound box by yourself? Thank you.