Java Deep Learning Projects

Real-time object detection using YOLO, JavaCV, and DL4J

In this project, we again used the transfer learning technique to solve another interesting problem, which is real-time object detection from video clips. We used a pre-trained YOLOv2 model, JavaCV, and DL4J libraries to solve this problem.

As stated in the chapter, we extended the image recognition idea to solve this problem. That is, our technique generates video frames as images and then recognizes objects from the frame using the bounding box approach. The takeaway is that although we used a video clip to show the evaluation, it still showed very good accuracy. And from the provided demo, anyone can observe that most of the objects in the clip were accurately identified. Thus, a ...