Counting people going in and out of one or multiple predetermined spaces to determine traffic flow and/or current and historic occupancy
The application of computer vision and deep learning have some remarkable outputs, such as object detection , object tracking and object counting etc. The combination of the mentioned use cases results in an application which allows the user to read a video and count the desired object in the desired direction. This has a phenomenal impact on retail and supply-chain sectors for various use-cases.
Where can this be used?
Applicable to areas like event entrances and exits, parking lots, rooms, delimited spaces, escalators, conveyor belts, roads, queues
What happens in the background?
The computer vision libraries allow us to read the video and process the images in the frame as required for further work.This can be fed from a live video feed or video file in RTSP format. These frames are then sent to the pre-trained object detection model which detects the desired object such as person,car etc. The object tracking model helps in following the detected objects and not to detect and count them again. Coordinate geometry and basic arithmetics are used to determine the direction of movement of the tracked objects so objects moving in the desired direction can be counted
How Padme helps?
Padme allows users to use it as a simple application for sophisticated tasks, earlier if the user had to train and retrain the model and then deploy it for his edge device ,padme can help them to do that just by a few mouse clicks on its user-friendly interface.The model is also trained and deployed on the application, the user just needs to select the model from the drop down list. Overall the user just needs to give an input video and they can observe the output clearly on the output screen.
- A new user needs to connects his edge device and provide us a video stream
- The user can draw polygons and triangle, polygons for determining the region of interest and triangles inside the polygon gives the direction
- The images in the video stream are processed and required objects are detected
- Unique id’s are assigned to the objects to track them with respect to the previous frame
- The ids which are passing in the mentioned directions are counted
- Finally in the output screen we can see number of visible objects, number of visible objects in various region of interests and count of objects passing the largest side of triangles within each polygon/roi
Impact on Retail and Supply chain Sectors.
The Ingress Egress model helps in automating or assisting the age-old practice of manual counting of products passing on the conveyor belt .This indeed helps in optimizing the process and resource management by giving an idea about necessary production based on the required demand levels. This can hugely impact the bottomline at the end of the financial year by regulating unnecessary production.