Detecting baggage threats is one of the most difficult tasks, even for expert\nofficers. Many researchers have developed computer-aided screening systems to\nrecognize these threats from the baggage X-ray scans. However, all of these\nframeworks are limited in identifying the contraband items under extreme\nocclusion. This paper presents a novel instance segmentation framework that\nutilizes trainable structure tensors to highlight the contours of the occluded\nand cluttered contraband items (by scanning multiple predominant orientations),\nwhile simultaneously suppressing the irrelevant baggage content. The proposed\nframework has been extensively tested on four publicly available X-ray datasets\nwhere it outperforms the state-of-the-art frameworks in terms of mean average\nprecision scores. Furthermore, to the best of our knowledge, it is the only\nframework that has been validated on combined grayscale and colored scans\nobtained from four different types of X-ray scanners.\n