MFMAN

Haze image recognition through multi-scale feature and multi-adversarial

This is the code of MFMAN

proposed a new diversity real haze dataset
dataset name Real-haze12,Real-clear12,Resources in Baidu Cloud，link：https://pan.baidu.com/s/1gZrmJsNhiXOWvKwb8pOzuQ :dnxd

Prerequisites:

Python3
PyTorch == 1.0.0 (with suitable CUDA and CuDNN version)
Numpy
argparse
PIL
tqdm

Baseline

·Deep residual learning(ResNet50)
·Domain Adaptive Neural Network(DANN)
·Deep Domain Confusion (DDC)
·Deep CORAL: Correlation Alignment for Deep Domain Adaptation (D-CORAL)
·Domain-Adversarial Training of Neural Networks (DANN-A)
the parametes is in the baseline\options.py,then run the train.py

MFMAN

fix the parameters in MFMAN\train.py,then run the train.py