Firearm and edged weapon detection in surveillance systems is critical for public safety. However, it faces significant challenges due to false positives. This study presents an automated data‐guided multi‐detector fusion framework with an innovative dataset balancing layer for model training. This balancing layer generates multiple datasets with carefully controlled image and instance distributions, reducing annotation bias and strengthening detector diversity. Several you only look once‐based detectors were trained independently on curated datasets. Their predictions are combined using weighted box fusion, which consistently outperforms other ensemble methods. The experimental results show a minimum false positive rate of 16%, the lowest among all tested datasets. This scalable and efficient approach presents a promising solution for weapon detection in safety‐critical environments. To ensure transparency and reproducibility, all datasets and source code related to this work are freely accessible at https://github.com/MuhammadIshtiaq/RQ2‐Dataset‐and‐Code . The repository provides detailed usage instructions, lists the required dependencies and describes the key algorithmic steps, allowing other researchers to reproduce the experiments and verify the findings.
23 July 2025 in UCP Journal of Engineering & Information Technology
The medium of instruction has a significant impact on effective communication and comprehension. The majority of literature is available in English because it is a universal language; however, if someone is provided with the information in their native language, they will comprehend the concept much more effectively. In computer science, source code of programming languages is written in the English language, whereas endemic language has its own impact. To address this gap, this study has rendered a framework, “Roman Urdu Programming Translator” (RUPT), that will be used to translate a program coded in Roman Urdu or Hindi into a proportionate C++ program. Actually, this will be an additional layer to the original C++ compiler, i.e., a programmer will write a program using special keywords of the Roman Urdu language, use RUPT to translate it into an equivalent C++ program, and then use the traditional C++ compiler to compile and run it on the computer. The special set of Roman Urdu keywords includes, e.g., “keyboardSay (ks)” instead of “cin”, “screenKiTaraf (skt)” instead of “cout”, “klea” instead of “for” etc. In this work, additional Urdu and Hindi Roman keywords are added to the original set of C++ language and will be replaced by the C++ equivalent keywords by our translator RUPT to convert it into a pure C++ program. RUPT is only composed of the lexical analysis phase. State of the art has increased the understanding and learning rate of novel users towards the field of computer science.
Object detection is central to public safety and surveillance, where reliable weapon identification is critical to mitigating security risks, and missed detections can have severe consequences. Existing ensemble methods using homogeneous architectures or standard fusion strategies do not adequately address occlusion, confidence unreliability, and contributor consensus, resulting in suboptimal recall and persistent false negatives. To address these limitations, this work presents an ensemble fusion framework combining convolutional neural network and transformer-based detectors for architectural diversity, together with a proposed Contributor Consensus Weighted Box Fusion (CC-WBF) method. The framework follows a three-stage approach: benchmarking for model selection, domain-specific pretraining, and task-focused finetuning. CC-WBF extends standard Weighted Box Fusion (WBF) through transitive clustering, weighted confidence aggregation, and contributor-aware rescaling to reduce the influence of weak or inconsistent predictions. Under matched cross-architecture detector combinations, CC-WBF increases mean recall from 0.9242 to 0.9434 on the Finetuning Dataset (FD) and from 0.7594 to 0.7871 on the Merged Military Dataset (MMD), corresponding to relative false negative rate reductions of 25.31% and 11.49%, respectively. These recall gains are statistically supported across matched detector combinations, but are accompanied by an increase in false discovery rate on both datasets. The results characterize CC-WBF as a recall-oriented fusion strategy, while the larger improvement of the complete framework reflects the combined contribution of architectural diversity and fusion. These findings support a statistically grounded, recall-oriented approach to weapon detection in surveillance, with implications for safety-critical computer vision. Implementation code and experimental resources are publicly available at https://github.com/MuhammadIshtiaq/RQ1-Dataset-and-Code.