search

matlab-deep-learning / Pretrained-Yolov9-Network-For-Object-Detection

YOLO v9 inference in MATLAB for Object Detection with yolov9t, yolov9s, yolov9m, yolov9c and yolov9e networks
https://in.mathworks.com/help/vision/ug/getting-started-with-object-detection-using-deep-learning.html
GNU Affero General Public License v3.0
6 stars 0 forks source link

model speed issue #1

Closed jihwan722 closed 1 month ago

jihwan722 commented 2 months ago

Hello,

I have been using the YOLOv8 model you uploaded previously and it has been working well. Currently, I am performing real-time object detection on images received from the camera in a simulation program. I understand that the YOLOv9 model is supposed to be more lightweight and faster compared to YOLOv8. However, after replacing the yolov8x.mat file with yolov9t.mat, I noticed that the detection speed is actually slower with YOLOv9t. Considering the model size, I would expect YOLOv9t to be much faster. Is there anything specific I should consider when applying the YOLOv9 model? The attached image shows the time taken for different YOLO versions when running the same scenario using the Simulink profiler. It can be seen that the YOLOv9t model takes more time compared to the YOLOv8x model.

Thank you. profiler_yolov8x profiler_yolov9t

zuberkhan175 commented 1 month ago

Hi,

The speed difference exists due to more ONNX operators in the YOLOv9 model (since imported via ONNX) compared to YOLOv8. We are willing to consider this as an enhancement in future.