RESUMEN
Introduction: Wearable assistive devices for the visually impaired whose technology is based on video camera devices represent a challenge in rapid evolution, where one of the main problems is to find computer vision algorithms that can be implemented in low-cost embedded devices. Objectives and Methods: This work presents a Tiny You Only Look Once architecture for pedestrian detection, which can be implemented in low-cost wearable devices as an alternative for the development of assistive technologies for the visually impaired. Results: The recall results of the proposed refined model represent an improvement of 71% working with four anchor boxes and 66% with six anchor boxes compared to the original model. The accuracy achieved on the same data set shows an increase of 14% and 25%, respectively. The F1 calculation shows a refinement of 57% and 55%. The average accuracy of the models achieved an improvement of 87% and 99%. The number of correctly detected objects was 3098 and 2892 for four and six anchor boxes, respectively, whose performance is better by 77% and 65% compared to the original, which correctly detected 1743 objects. Discussion: Finally, the model was optimized for the Jetson Nano embedded system, a case study for low-power embedded devices, and in a desktop computer. In both cases, the graphics processing unit (GPU) and central processing unit were tested, and a documented comparison of solutions aimed at serving visually impaired people was performed. Conclusion: We performed the desktop tests with a RTX 2070S graphics card, and the image processing took about 2.8 ms. The Jetson Nano board could process an image in about 110 ms, offering the opportunity to generate alert notification procedures in support of visually impaired mobility.
RESUMEN
Passive content fingerprinting is widely used for video content identification and monitoring. However, many challenges remain unsolved especially for partial-copies detection. The main challenge is to find the right balance between the computational cost of fingerprint extraction and fingerprint dimension, without compromising detection performance against various attacks (robustness). Fast video detection performance is desirable in several modern applications, for instance, in those where video detection involves the use of large video databases or in applications requiring real-time video detection of partial copies, a process whose difficulty increases when videos suffer severe transformations. In this context, conventional fingerprinting methods are not fully suitable to cope with the attacks and transformations mentioned before, either because the robustness of these methods is not enough or because their execution time is very high, where the time bottleneck is commonly found in the fingerprint extraction and matching operations. Motivated by these issues, in this work we propose a content fingerprinting method based on the extraction of a set of independent binary global and local fingerprints. Although these features are robust against common video transformations, their combination is more discriminant against severe video transformations such as signal processing attacks, geometric transformations and temporal and spatial desynchronization. Additionally, we use an efficient multilevel filtering system accelerating the processes of fingerprint extraction and matching. This multilevel filtering system helps to rapidly identify potential similar video copies upon which the fingerprint process is carried out only, thus saving computational time. We tested with datasets of real copied videos, and the results show how our method outperforms state-of-the-art methods regarding detection scores. Furthermore, the granularity of our method makes it suitable for partial-copy detection; that is, by processing only short segments of 1 second length.