Color comprises one of the most significant and dominant cues for various applications. As one of the most noticeable and stable attributes of vehicles, color can constitute a valuable key component in several practices of intelligent surveillance systems. In this paper, we propose a deep-learning-based framework that combines semantic segmentation masks with pixels clustering for automatic vehicle color recognition. Different from conventional methods, which usually consider only the features of the vehicle's front side, the proposed algorithm is able for view-independent color identification, which is more effective for the surveillance tasks. To the best of our knowledge, this is the first work that employs semantic segmentation masks along with color clustering for the extraction of the vehicle's color representative parts and the recognition of the dominant color, respectively. To evaluate the performance of the proposed method, we introduce a challenging multi-view dataset of 500 car-related RGB images extending the publicly available DSMLR Car Parts dataset for vehicle parts segmentation. The experiments demonstrate that the proposed approach achieves excellent performance and accurate results reaching an accuracy of 93.06% in the multi-view scenario. To facilitate further research, the evaluation dataset and the pre-trained models will be released at https://github.com/klearchos-stav/vehicle_color_recognition.

Gas cylinder detection and the identification of their characteristics hold considerable potential for enhancing safety and operational efficiency in several applications, including industrial and warehouse operations. These tasks gain significance with the growth of online trade, emerging as critical instruments to combat environmental crimes associated with hazardous substances’ illegal commerce. However, the lack of relevant datasets hinders the effective utilization of deep learning techniques within this domain. In this study, we introduce CylinDeRS, a domain-specific dataset for gas cylinder detection and the classification of their attributes in real-world scenes. CylinDeRS contains 7060 RGB images, depicting various challenging environments and featuring over 25,250 annotated instances. It addresses two tasks: (a) the detection of gas cylinders as objects of interest, and (b) the attribute classification of the detected gas cylinder objects for material, size, and orientation. Extensive experiments using state-of-the-art (SotA) models are reported to validate the dataset’s significance and application prospects, providing baselines for further performance evaluation and in-depth analysis. The results show a maximum mAP of 91% for the gas cylinder detection task and a maximum accuracy of 71.6% for the attribute classification task, highlighting the challenges posed by real-world scenarios and underlining the proposed dataset’s importance in advancing the field.