search

fizyr / keras-retinanet

Keras implementation of RetinaNet object detection.
Apache License 2.0
4.38k stars 1.96k forks source link

Regression loss keeps zero #1489

Closed KevinZhoutianyi closed 3 years ago

KevinZhoutianyi commented 4 years ago

Console

5/8 [=================>............] - ETA: 2s - loss: 2.4863e-08 - regression_loss: 0.0000e+00 - classification_loss: 2.46/8 [=====================>........] - ETA: 1s - loss: 2.4947e-08 - regression_loss: 0.0000e+00 - classification_loss: 2.47/8 [=========================>....] - ETA: 0s - loss: 2.4671e-08 - regression_loss: 0.0000e+00 - classification_loss: 2.48/8 [==============================] - ETA: 0s - loss: 2.4837e-08 - regression_loss: 0.0000e+00 - classification_loss: 2.4837e-08

data

/data/zhoutianyi/keras-retinanet/custom_dataset/test/1_1_2_0_1.png,405,63,445,97,rivet /data/zhoutianyi/keras-retinanet/custom_dataset/test/1_1_2_0_1.png,483,63,522,96,rivet /data/zhoutianyi/keras-retinanet/custom_dataset/test/1_1_2_0_1.png,559,63,593,93,rivet ...

command

python3 keras_retinanet/bin/train.py --steps 8 csv train_rivets.csv classmapping.csv

extra

when i use the trained h5 inference model, it cant predict. The score is -1, and there's no any bounding box shown on the img

stale[bot] commented 3 years ago

This issue has been automatically marked as stale due to the lack of recent activity. It will be closed if no further activity occurs. Thank you for your contributions.