People detection using Yolo V5 on Flo Edge One

The Yolo V5 object detection model is a powerful tool that can detect a wide variety of labels, including people, cars, and more. In this case study, we will focus on the people detection aspect of the model and analyze its performance.

Data and Model

We were provided with footage from a cricket match, and the Yolo V5 model was trained on the Coco database, which contains around 90 different labels. The model has an inference time of 40 to 50 milliseconds per frame, resulting in a smooth 20 FPS output. The model is capable of detecting people with 91% confidence.

Performance Analysis:

During the analysis of the provided footage, we observed that the model was able to detect people accurately. The detection score increased from 50% to 90% as soon as the person’s entire body was in the frame. This shows that the model can accurately identify people, even in complex scenes.

Moreover, the Yolo V5 model is impressive because it can detect a wide range of labels with the same architecture. The model runs entirely on the Flo Edge One GPU, leaving most of the CPU space free. This means that users can run a CPU-oriented process in the background without any trouble.

Conclusion:

The Yolo V5 object detection model is a powerful tool that can detect people with high accuracy. Its ability to detect a wide range of labels with the same architecture and run on a low-power GPU is impressive. This model can be used in various applications, including surveillance, security, and even sports analysis.