Open ahn19 opened 11 months ago
I ran the following code against the appa face dataset (https://chalearnlap.cvc.uab.cat/dataset/26/description/) https://github.com/yu4u/age-gender-estimation/blob/master/age_estimation/train.py
However, as you can see from the results below, the performance results are different in Debug mode and Release mode (I am using PyCharm). The results in Debug mode are better than those in Release mode.
Why is that so?
release mode result:
Epoch 99/100 64/64 [==============================] - ETA: 0s - loss: 3.8237 - age_mae: 9.8588 Epoch 99: val_age_mae improved from 10.31055 to 10.26129, saving model to E:\Project\ref_code\face2age_age-gender-estimation-master\age_estimation\..\checkpoints\weights.099-3.875-10.261.hdf5 64/64 [==============================] - 63s 986ms/step - loss: 3.8237 - age_mae: 9.8588 - val_loss: 3.8748 - val_age_mae: 10.2613 - lr: 8.0000e-04 Epoch 100/100 64/64 [==============================] - ETA: 0s - loss: 3.8306 - age_mae: 9.9421 Epoch 100: val_age_mae did not improve from 10.26129 64/64 [==============================] - 63s 975ms/step - loss: 3.8306 - age_mae: 9.9421 - val_loss: 3.8752 - val_age_mae: 10.2669 - lr: 8.0000e-04 Process finished with exit code 0
debug mode result:
Epoch 98/100 64/64 [==============================] - ETA: 0s - loss: 2.7404 - age_mae: 3.1714 Epoch 98: val_age_mae did not improve from 5.98561 64/64 [==============================] - 61s 945ms/step - loss: 2.7404 - age_mae: 3.1714 - val_loss: 3.6023 - val_age_mae: 6.1649 - lr: 8.0000e-04 Epoch 99/100 64/64 [==============================] - ETA: 0s - loss: 2.7362 - age_mae: 3.1317 Epoch 99: val_age_mae did not improve from 5.98561 64/64 [==============================] - 61s 955ms/step - loss: 2.7362 - age_mae: 3.1317 - val_loss: 3.5799 - val_age_mae: 6.1791 - lr: 8.0000e-04 Epoch 100/100 64/64 [==============================] - ETA: 0s - loss: 2.7008 - age_mae: 3.0804 Epoch 100: val_age_mae did not improve from 5.98561 64/64 [==============================] - 62s 956ms/step - loss: 2.7008 - age_mae: 3.0804 - val_loss: 3.5764 - val_age_mae: 6.0456 - lr: 8.0000e-04 Process finished with exit code 0
I ran the following code against the appa face dataset (https://chalearnlap.cvc.uab.cat/dataset/26/description/) https://github.com/yu4u/age-gender-estimation/blob/master/age_estimation/train.py
However, as you can see from the results below, the performance results are different in Debug mode and Release mode (I am using PyCharm). The results in Debug mode are better than those in Release mode.
Why is that so?
release mode result:
debug mode result: