from sklearn.ensemble import RandomForestClassifier
from utils import *All custom model must have methods:
- fit(...)
- predict(...)
You can add custom features like what I do below by making new class which inherited some class.
If your model doesn't have the required methods, create custom ones for it.
class MyRandomForestClf(RandomForestClassifier):
def __init__(self, n_estimators_added_before_each_fit, **kwargs):
self.n_estimators_added_before_each_fit = n_estimators_added_before_each_fit
super(MyRandomForestClf, self).__init__(**kwargs)
def fit(self, x, y, **kwargs):
self.n_estimators += self.n_estimators_added_before_each_fit
super(MyRandomForestClf, self).fit(x, y, **kwargs)To use the following tutorial, please set the "attributes" array in configure yaml file contains string "hp".
init_model = MyRandomForestClf(n_estimators_added_before_each_fit=1)Call this function, you'll get a dict of: key -> value: str(attribute) -> MyModel
attr_models = read_config_and_train(model=init_model, img_dir='data/images')MyModel.save_model(filename)
Note: Not sure whether using this function to save model that is not from sklearn is appropriate.
example:
attr_models['hp'].save_model('filename_you_want')- Existing MyModel load other one:
MyModel.load_model(file_path='filename_you_want') - Creating new MyModel object:
MyModel(model_path_to_load='model_name_you_want_to_load')
Note: Not sure whether using this function to save model that is not from sklearn is appropriate.
example:
- For existing MyModel:
attr_models['hp'].load_model('filename_you_want') # It'll cover your old model- For creating MyModel:
new_my_model = MyModel(model_path_to_load='filename_you_want') MyModel.predict_one_image(
filename= ???(str),
show_original_img= ???(bool),
show_preprocessed_img= ???(bool,
)example:
pred = attr_models['hp'].predict_one_image(filename=filename,
show_original_img=True,
show_preprocessed_img=True)
print(pred)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent