Vision Seeker: Deep Learning based Missing Person Detection and Recognition in the CCTV and Prerecorded Video

The increasing number of missing persons cases which is a serious challenge faced by all of the world. Despite of having CCTV in numerous places, there is a lack of technological methods to find them through it in an efficient way. With the ever-increasing population in metropolitan cities, finding a person in a crowded area is also a tedious task to do. This paper proposes Vision seeker- an AI which will help the law and enforcement to find the missing person in a short time. The system integrates three pretrained models i.e. GFP- GAN -which will help to increase the quality of the image, secondly MTCNN for detecting the image even in conditions like poor lighting, crowdy areas, etc. from the video by checking it through all frames, and lastly face net with the InceptionResnetV1 architecture, implemented using the facenet_pytorch library will recognize the facial structure by checking the facial embeddings and will compare it with the database of that individual. This model will help to integrate this with the real time access of the CCTV and in return will give us the image and also the timestamps on which the missing person is found. It will also show the location in which that individual has been found. The primary contribution lies in the novel system-level integration and practical deployment pipeline, rather than algorithmic innovation. By implementing this solution, various persons can be located in quick amount of time which will ensure the public safety and will improve the technological implementations of the surveillance and shows us the immense power of the AI in solving real world problems. The system giving the accuracy of 97.63% in pre-recorded video and 87.1% accuracy in real time.