Deep Learning Interpretability

Save Model

The templates are saved in mongodb database for future explanation, the template download URL is the key to access in the future.

model: Keras model URL saved in file type H5

Example

URL : /api/uploadmodel

Method : POST

Data example

{
    "model": "https://github.com/viniaraujoo/model_Example/raw/master/model_incep.h5"
}

LIME:

model: Template key (URL) saved to the database

example: URL of the image you want to use as an example of model explanation

top_labels: Explanation for K labels with higher probability.

num_samples: size of the neighborhood to learn the linear model

hide_color: 0 or 1

Example

URL : /api/explanationlime

Method : GET

Data example

{
    "model": "https://github.com/viniaraujoo/model_Example/raw/master/model_incep.h5",
    "example": "https://raw.githubusercontent.com/marcotcr/lime/master/doc/notebooks/data/cat_mouse.jpg",
    "top_labels": 5,
    "num_samples": 1000,
    "hide_color": 0

}

Success Response

Code : 200 OK

Error Response

Code : 400

SHAP:

model: Template key (URL) saved to the database

example: Image URL in numpy compressed ZIP file (Image that you want the explanation).

train: ZIP of the set of images numpy array train format of the explanation.

Example

URL : /api/shap

Method : GET

Data example

{
    "model": "https://github.com/viniaraujoo/model_Example/raw/master/model_incep.h5",
    "example": "",
    "train": ,

}

Success Response

Code : 200 OK

Error Response

Code : 400

License