4 classes of weather have been predicted with using CNN from the scratch.

• HAZE
• SUNNY
• SNOWY
• RAINY

• Apply Transfer Learning via feature extraction
• Apply Transfer Learning via fine-tuning

This program is built using the following versions: Python 3.7 Tensorflow 2.0.0 Keras 2.3.1

Install the dependencies.

$ pip install keras
$ pip install tensorflow
$ pip install numpy,pandas,pickle
$ pip install opencv-python
$ pip install imutils

For training the model

$ python train.py --dataset dataset --model weather.model --labelbin lb.pickle

For testing the blind dataset

$ python blindtest.py --model weather.model --labelbin lb.pickle --image examples/test

For testing only one photo while seeing the results in OpenCV

$ python classify.py --model weather.model --labelbin lb.pickle --image examples/HAZE_1.png

MIT