University of Twente Student Theses


Real-time detection of photographing and filming on embedded systems

Ismayilov, R.H.O. (2022) Real-time detection of photographing and filming on embedded systems.

[img] PDF
Abstract:Abstract—Current developments in image and video capturing technologies enable the possibility of non-consensual distribution of one’s identity information. Whether it is accidental photography or a deliberate attempt of filming, with emerging privacy concerns, potential methods of preventing a person from being identified by most facial recognition systems are currently being investigated. Several wearable solutions, such as jewelry or glasses, targeted to prevent facial identification exist; however, most require manual control due to their passive nature. This paper solves this problem with an embedded system capable of automatic real-time inhibition of photography and filming. Using a lightweight YoloFastestV2 deep learning model in combination with NCNN and MNN inference frameworks, targeted for optimal performance on embedded devices, an object detection algorithm is trained and deployed on Raspberry Pi devices to identify when the user is being filmed. Glasses with variable lens transparency are used in the system and instructed to turn dark when filming is detected to prevent identity recognition. Precision-recall curves are used as a metric to evaluate the designed models, and differences between NCNN and MNN frameworks are examined. Based on the results, the proposed system achieves an accuracy of 89% when evaluated using images depicting expected filming scenarios. Real-world experiments are also conducted to validate the performance, and results demonstrate that accurate detections with an inference time of 46ms are achievable on Raspberry Pi 4.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Programme:Electrical Engineering BSc (56953)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page