bing0037 / keras-yolo3 Goto Github PK

Epoch 38/50
4/4 [==============================] - 10s 2s/step - loss: 92487.3184 - val_loss: 75600.4219
Epoch 39/50
4/4 [==============================] - 10s 3s/step - loss: 89534.9199 - val_loss: 63017.3125
Epoch 40/50

I trained my model on darknet on my dataset.I have successfully converted my model in keras but when i testing this keras model i am getting many small bounding box but the labels are correct.

How can I train yolo on grayscale images? I have changed channels to 1 in .cfg file as well as Input(shape=(None, None, 1)) in train.py. I got this error: ValueError: Error when checking input: expected input_1 to have shape (None, None, 1) but got array with shape (416, 416, 3) How to preprocess images properly?

In training raccoon data, having the error:

Epoch 00094: ReduceLROnPlateau reducing learning rate to 9.999999939225292e-10.
Epoch 00094: early stopping.

how to solve the problem?

I want to increase the amount of anchor to detect more exactly. But i can not find out it. Please help me

HI,

i would like to know if there is a way to perform prediction on new images or videos and then crop and save only the bounding box

thank you