def features():
response = {
'id': em.data.index.name or 'index',
'label': em.label_name,
'features': list(em.default_data.keys())
}
return jsonify(response)
def sample_table():
#N = request.args.get('N', 10, int)
sample_ids = random.sample(range(1, len(em.data)), 10)
sample_table = []
for ids in sample_ids:
sample_table.append(explainid_non_flask(str(ids)))
return jsonify(sample_table)
def explain():
data = dict(ChainMap(request.args, em.default_data))
data = em.cast_dct(data)
probability, explanation = em.explain_dct(data)
return jsonify({'data': dict(data),
'probability': probability,
'explanation': explanation})
def predict(args):
#data = dict(ChainMap(request.args, em.default_data))
data = dict(ChainMap(args, em.default_data))
data = em.cast_dct(data)
probability, explanation = em.explain_dct(data)
return jsonify({'probability': probability})
#test
#x={"StreamingTV":"No","MonthlyCharges":70.35,"PhoneService":"No","PaperlessBilling":"No","Partner":"No","OnlineBackup":"No","gender":"Female","Contract":"Month-to-month","TotalCharges":1397.475,"StreamingMovies":"No","DeviceProtection":"No","PaymentMethod":"Bank transfer (automatic)","tenure":29,"Dependents":"No","OnlineSecurity":"No","MultipleLines":"No","InternetService":"DSL","SeniorCitizen":"No","TechSupport":"No"}
#predict(x)
def predict(args):
im = Image.open(BytesIO(base64.b64decode(args['image'][22:])))
im_array = np.asarray(im)
print(im_array.shape)
im_float = im_array.astype('float32')
img_tensor = np.expand_dims(im_float, axis=0)
im_normalised = (img_tensor / 256)
activations = activation_model.predict(im_normalised)
result_layer = activations[-1]
result = result_layer[0, :]
return jsonify({'prediction': result})
def predict(args):
im = Image.open(BytesIO(base64.b64decode(args['image'][22:])))
im_blur = im.filter(ImageFilter.GaussianBlur(5))
im_small = im_blur.resize((28, 28), Image.ANTIALIAS)
im_bitmap = im_small.convert("L")
im_array = np.asarray(im_bitmap)
im_float = im_array.astype('float32')
im_normalised = (im_float / 256)
im_tensor = torch.tensor(im_normalised)
im_tensor_shaped = im_tensor.reshape(1, 1, 28, 28)
with torch.no_grad():
logps = model(im_tensor_shaped)
ps = torch.exp(logps)
probab = list(ps.cpu().numpy()[0])
predict_val = probab.index(max(probab))
return jsonify({'prediction': predict_val})
def explain(args):
data = dict(ChainMap(args, em.default_data))
data = em.cast_dct(data)
probability, class_pred, explanation = em.explain_dct_class(data)
return jsonify({'data': dict(data),
'quality_prediction': class_pred,
'explanation': explanation})
#test
#Poor_args={"fixedAcidity":7.4, "volatileAcidity":0.7, "citricAcid":0, "residualSugar":1.9, "chlorides":0.076, "freeSulfurDioxide":11, "totalSulfurDioxide":34, "density":0.9978, "pH":3.51, "sulphates":0.56, "Alcohol":9.4}
#Excellent_args={"fixedAcidity":5.9, "volatileAcidity":0.550, "citricAcid":0.10, "residualSugar":2.2, "chlorides":0.062, "freeSulfurDioxide":39.0, "totalSulfurDioxide":51.0, "density":0.99512, "pH":3.52, "sulphates":0.76, "Alcohol":11.2}
#predict(Poor_args)
#predict(Excellent_args)
#explain(Poor_args)
#explain(Excellent_args)
def modelname():
return jsonify({'modelname': em.model_name})
def default():
return jsonify(em.default_data)
def size():
return jsonify({'size': len(em.data)})
def stats():
return jsonify(em.stats)
def categories():
return jsonify(
{feat: dict(enumerate(cats))
for feat, cats in em.categories.items()})
def dataset():
return jsonify({'dataset': em.dataset})
def sample_table():
sample_ids = random.sample(range(1,len(em.data)),10)
sample_table = []
for ids in sample_ids:
sample_table.append(explainid(str(ids)))
return jsonify(sample_table)
def predict(args):
data = dict(ChainMap(args, em.default_data))
data = em.cast_dct(data)
probability, class_pred, explanation = em.explain_dct_class(data)
return jsonify({'quality_prediction': class_pred})