search

hunglc007 / tensorflow-yolov4-tflite

YOLOv4, YOLOv4-tiny, YOLOv3, YOLOv3-tiny Implemented in Tensorflow 2.0, Android. Convert YOLO v4 .weights tensorflow, tensorrt and tflite
https://github.com/hunglc007/tensorflow-yolov4-tflite
MIT License
2.25k stars 1.24k forks source link

Performance drop #415

Open zz100chan opened 2 years ago

zz100chan commented 2 years ago

Hi everyone, I have been working to deploy my YOLOv4 model to Flask recently. When I finished the deployment, I checked the performance of detection between the tf-yolov4 model and the darknet model performance. With IOU and confidence set to 0.5, the tf-yolov4 was performing way worse than the darknet model.

Let alone the tf-yolov4 can only detect the maximum of 50 objects in the pic but darknet can do more than that, the classes of detection is also different from the darknet model.

Does anyone encounter the same situation? Any insight or solution is very appreciated! Thank you very much!

ghost commented 2 years ago

I got the worse result than darknet as well and didn't get solution yet.