This paper demonstrates the framework and results from the team “Florida International University - University of Miami (FIU-UM)” in the TRECVID 2017 [1] Ad-hoc Video Search (AVS) task [2]. The following four runs were submitted: • M D FIU UM.17 1: CNN features + Linear SVM • M D FIU UM.17 2: CNN features + Linear SVM + Scores from other groups • M D FIU UM.17 3: CNN features + Linear SVM + Rectified Linear Score Normalization • M D FIU UM.17 4: CNN features + Linear SVM + Scores from other groups + Rectified Linear Score Normalization In the first step, the features are extracted by the convolutional neural network (CNN) structure of GoogLeNet [3] for all runs. Then, the scores of each concept for the key frames are generated using the linear support vector machine (SVM) classifiers. For Run 2 and Run 4, the scores generated by the ITI-CERTH team last year are integrated for score fusion and enhancement. Meanwhile, for Run 3 and Run 4, a new rectified linear score normalization algorithm is used. As a result, from the submission results, Run 2 outperforms the other three runs. The submission details are listed as follows. • Class: M (Manually-assisted runs) • Training type: D (IACC & non-IACC non-TRECVID data) • Team ID: FIU-UM (Florida International University - University of Miami) • Year: 2017