Discrepancies in Object Detection Results between Ultralytics and DeepStream at Yolov8

[1712catfish]

Hi,

I hope this message finds you well. I am reaching out to seek your expertise and assistance regarding an issue I am currently facing with the conversion of the YOLOv8 model to the FP32 engine. My objective is to utilize this model within DeepStream for primary inference (pgie), specifically for detecting a single person class. However, I have encountered significant discrepancies in the detection results when comparing the outcomes derived from Ultralytics with those obtained from DeepStream.

To provide a detailed account of the challenges I am experiencing, I have documented the steps taken, the expected versus actual results, and any troubleshooting attempts thus far on the NVIDIA Developer Forums. You can find the complete post at the following link: https://forums.developer.nvidia.com/t/discrepancies-in-object-detection-results-between-ultralytics-and-deepstream-at-yolov8/282021.

As of now, I have not received a response to my forum post, and I am keenly seeking your guidance to understand and resolve this issue. Your insights would be invaluable to me in navigating this challenge and achieving accurate and consistent detection results.

I eagerly await your expert advice and thank you in advance for your time and assistance. Please let me know if you require any further information or clarification regarding my inquiry.