You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I would like to ask you about some strange behavior I am experiencing with Yolov5.
I'm using a Yolov5s model for my object detection problem on a custom dataset. It's a small dataset of about 300 images with a resolution of 1920 × 1080 pixels. The dataset consists of 4 classes, with approximately 100 instances per class. The model achieves a decent mAP of 0.55 at a threshold of 0.5.
I read that using a Yolov5 P6 model is better for images with resolutions larger than 640. Therefore, the next step was to try a larger network, specifically the Yolov5s6, in hopes of achieving better performance. Unfortunately, the performance deteriorated drastically, as shown in the following image:
Both models were trained using the standard hyperparameters specified in the yolov5/data/hyps/hyp.scratch-low.yaml file.
Do you have any hints or suggestions on how to solve this problem?
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Hi everyone!
I would like to ask you about some strange behavior I am experiencing with Yolov5.
I'm using a Yolov5s model for my object detection problem on a custom dataset. It's a small dataset of about 300 images with a resolution of 1920 × 1080 pixels. The dataset consists of 4 classes, with approximately 100 instances per class. The model achieves a decent mAP of 0.55 at a threshold of 0.5.
I read that using a Yolov5 P6 model is better for images with resolutions larger than 640. Therefore, the next step was to try a larger network, specifically the Yolov5s6, in hopes of achieving better performance. Unfortunately, the performance deteriorated drastically, as shown in the following image:
Both models were trained using the standard hyperparameters specified in the
yolov5/data/hyps/hyp.scratch-low.yaml
file.Do you have any hints or suggestions on how to solve this problem?
Beta Was this translation helpful? Give feedback.
All reactions