DENets

Defocus blur detection via boosting diversity of deep ensemble networks

We have two networks: EFENet and AENet

Requirement:

python==3.6

pytorch==1.1.0

tensorflow==1.5.0

Test_datasets and Results:

CUHK: Contains 100 testing images of CUHK Dataset and it's GT.

DUT: Contains 500 testing images of DUT Dataset and it's GT.

Test_dataset:

Baidu Netdisk link: https://pan.baidu.com/s/1Uj-i2CJVwt-Tz_d4NdDtEw password: y5kv

Google Drive link: https://drive.google.com/file/d/109c36vRDagDcjUbUtBouausXz0DOkA-w/view?usp=sharing

Results:

Baidu Netdisk link: https://pan.baidu.com/s/1UMN2neErD1G7i889pRhNOA password: 977n

Google Drive link: https://drive.google.com/file/d/1XuiccdbVrB5vUtqHVFCg8Z4US66qXyoo/view?usp=sharing

Parameters:

EFENet:

Baidu Netdisk link: https://pan.baidu.com/s/1BK0B4JNGEtTEXpQgL30vgA password: 3fm0

Google Drive link: https://drive.google.com/file/d/1Ug67nYEBNYYsHRuMlQLz9klxCQNKJIl7/view?usp=sharing

AENet:

Baidu Netdisk link: https://pan.baidu.com/s/1BkI6ufuaJBQafJYfPmDQwQ password: ufpq

Google Drive link: https://drive.google.com/file/d/1LJ_pBAGx6qDD7cboSm1BKuRFUevi38Wu/view?usp=sharing

Test:

You can run EFENet/test.py to test EFENet and run AENet/src/test.py to test AENet.