Shoplifting Prevention
Detecting shoplifters on shop surveillance cameras by recognizing the patterns in human behavior.
Detecting shoplifters on shop surveillance cameras by recognizing the patterns in human behavior.
Shoplift is a persisting social and criminal issue that costs retail business owners significant sums of money over the course of their activities. Identifying, pursuing and keeping perpetrators accountable is a current way to handle the issue. However, a more efficient way is to spot shoplifters beforehand to prevent thefts.
SOLVVE team took on the development of such prevention tools using machine learning to detect shoplifting on existing systems of surveillance cameras instead of high-resolution ones.
The key point is using video footage from surveillance cameras in shops, supermarkets, trade centers, and other public facilities to detect the anomalies in the human behavior and classify anomalies as shoplifting, robbery, stealing, etc. as compared to regular shoppers’ behavior.
SOLVVE team used CNN i3d with transfer learning and fine-tuning techniques along with UCFCrime dataset of videos catching anomalies in human behavior to train the model that can detect such anomalies in real time and send signals about the shoplifting actions to those in charge of security and crime prevention.
Computer vision, Instance segmentation, CNN; Tensorflow, Keras, GluonCV, OpenCV, Pytorch, Numpy