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 have gotten predicted results(images) with pale purple spot, which presented randomly when using the pre-trained model named "InceptionResNet-v2" or "fpn_inception.h5"(download).
In addition, the phenomenon also exists when using the model trained on my own dataset. I set ['model']['g_name'] "fpn_inception" in config/config.yaml and used other default settings when training.
Furthermore, when using “fpn_mobilenet”, there is no such problem.
Anyone who has found the cause of this problem would be most grateful for your sharing!Thanks.
The text was updated successfully, but these errors were encountered:
I also met this issue. #18 (comment) @KupynOrest@t-martyniuk said it was bug in the inference and updated the pre-trained model, but seems still occurs on other testing images.
I have gotten predicted results(images) with pale purple spot, which presented randomly when using the pre-trained model named "InceptionResNet-v2" or "fpn_inception.h5"(download).
In addition, the phenomenon also exists when using the model trained on my own dataset. I set ['model']['g_name'] "fpn_inception" in config/config.yaml and used other default settings when training.
Furthermore, when using “fpn_mobilenet”, there is no such problem.
Anyone who has found the cause of this problem would be most grateful for your sharing!Thanks.
The text was updated successfully, but these errors were encountered: