Abnormal activity detection is one of the most challenging tasks in the field\nof computer vision. This study is motivated by the recent state-of-art work of\nabnormal activity detection, which utilizes both abnormal and normal videos in\nlearning abnormalities with the help of multiple instance learning by providing\nthe data with video-level information. In the absence of temporal-annotations,\nsuch a model is prone to give a false alarm while detecting the abnormalities.\nFor this reason, in this paper, we focus on the task of minimizing the false\nalarm rate while performing an abnormal activity detection task. The mitigation\nof these false alarms and recent advancement of 3D deep neural network in video\naction recognition task collectively give us motivation to exploit the 3D\nResNet in our proposed method, which helps to extract spatial-temporal features\nfrom the videos. Afterwards, using these features and deep multiple instance\nlearning along with the proposed ranking loss, our model learns to predict the\nabnormality score at the video segment level. Therefore, our proposed method 3D\ndeep Multiple Instance Learning with ResNet (MILR) along with the new proposed\nranking loss function achieves the best performance on the UCF-Crime benchmark\ndataset, as compared to other state-of-art methods. The effectiveness of our\nproposed method is demonstrated on the UCF-Crime dataset.\n