Base Epoch Test 1
Purpose
- Select base models from images
model
- base network : FPN
- encoder : resnet34
- pretrained : imagenet
I compare the epoch images of the model to determine the best epoch to conduct the experiment.
- The validataion dataset consists of 100 images.
- Each line consists of 10 images.
- images are created every three epochs.
line 1
Image
line 2
Image
line 3
Image
line 4
Image
line 5
Image
line 6
Image
line 7
Image
line 8
Image
line 9
Image
line 10
Image
-
Comparing images of the model, we can confirm that the prediction results of the images of epoch3 and epoch6 are different.
-
However, for images after epoch6, the prediction results are the same. Therefore, we judge that progressing epoch6 learning has progressed sufficiently.
-
Afterwards, the image between epoch1 and 6 will be compared again to determine the best epoch.
Base Epoch Test 1
PurposemodelI compare the epoch images of the model to determine the best epoch to conduct the experiment.
line 1
Image
line 2
Image
line 3
Image
line 4
Image
line 5
Image
line 6
Image
line 7
Image
line 8
Image
line 9
Image
line 10
Image
Comparing images of the model, we can confirm that the prediction results of the images of epoch3 and epoch6 are different.
However, for images after epoch6, the prediction results are the same. Therefore, we judge that progressing epoch6 learning has progressed sufficiently.
Afterwards, the image between epoch1 and 6 will be compared again to determine the best epoch.