def demo_prediction():
"""Route to a jsonified version of deep learning model predictions, for
demo case
Returns
-------
dict
Deep learning model prediction for demo page (depends on the chosen
model, either feature detection or semantic segmentation)
"""
filename = request.args.get('img')
filename = os.path.join("deeposlandia", filename[1:])
dataset = request.args.get('dataset')
agg_value = dataset == "mapillary"
model = request.args.get('model')
utils.logger.info("file: {}, dataset: {}, model: {}".format(
filename, dataset, model))
if model == "feature_detection":
predictions = predict([filename], dataset, model, aggregate=agg_value)
return jsonify(predictions)
elif model == "semantic_segmentation":
predictions = predict([filename],
dataset,
model,
aggregate=agg_value,
output_dir=PREDICT_FOLDER)
return jsonify(predictions)
else:
utils.logger.error(("Unknown model. Please choose "
"'feature_detection' or 'semantic_segmentation'."))
return ""
def predictor_demo(model, dataset, image):
"""Route to a jsonified version of deep learning model predictions, for
demo case
Parameters
----------
model : str
Considered research problem
dataset : str
Considered dataset
image : str
Name of the demo image onto the server
Returns
-------
dict
Deep learning model prediction for demo page (depends on the chosen
model, either feature detection or semantic segmentation)
"""
logger.info("file: %s, dataset: %s, model: %s", image, dataset, model)
image_info = recover_image_info(dataset, image)
predictions = predict(
[os.path.join(app.static_folder, image_info["image_file"])],
dataset,
model,
output_dir=PREDICT_FOLDER,
)
if model == "featdet":
predicted_image = "sample_image/prediction.png"
predicted_labels = predictions[
os.path.join(app.static_folder, image_info["image_file"])
]
elif model == "semseg":
predicted_image = os.path.join(
"predicted", image_info["label_file"].split("/")[-1]
)
predicted_labels = predictions["labels"]
else:
raise ValueError(
(
"Unknown model, please choose amongst %s.", AVAILABLE_MODELS
)
)
return render_template(
dataset + "_demo.html",
model=model,
image_filename=image_info["image_file"],
label_filename=image_info["label_file"],
ground_truth_labels=image_info["labels"],
predicted_filename=predicted_image,
predicted_labels=predicted_labels,
)
def predictor_demo(model, dataset, image):
"""Route to a jsonified version of deep learning model predictions, for
demo case
Parameters
----------
model : str
Considered research problem (either `feature_detection` or
`semantic_segmentation`)
dataset : str
Considered dataset (either `shapes` or `mapillary`)
image : str
Name of the demo image onto the server
Returns
-------
dict
Deep learning model prediction for demo page (depends on the chosen
model, either feature detection or semantic segmentation)
"""
agg_value = dataset == "mapillary"
logger.info("file: %s, dataset: %s, model: %s", image, dataset, model)
agg_value = dataset == "mapillary"
image_info = recover_image_info(dataset, image)
predictions = predict(
[os.path.join(app.static_folder, image_info["image_file"])],
dataset,
model,
aggregate=agg_value,
output_dir=PREDICT_FOLDER,
)
if model == "feature_detection":
predicted_image = "sample_image/prediction.png"
predicted_labels = predictions[os.path.join(app.static_folder,
image_info["image_file"])]
elif model == "semantic_segmentation":
predicted_image = os.path.join("predicted",
image_info["label_file"].split("/")[-1])
predicted_labels = predictions["labels"]
else:
raise ValueError(("Unknown model, please provide 'feature_detection'"
"or 'semantic_segmentation'."))
return render_template(
dataset + "_demo.html",
model=model,
image_filename=image_info["image_file"],
label_filename=image_info["label_file"],
ground_truth_labels=image_info["labels"],
predicted_filename=predicted_image,
predicted_labels=predicted_labels,
)
def prediction():
"""Route to a jsonified version of deep learning model predictions, for
client tool
Returns
-------
dict
Deep learning model predictions
"""
filename = os.path.basename(request.args.get('img'))
filename = os.path.join(app.config['UPLOAD_FOLDER'], filename)
dataset = request.args.get('dataset')
model = request.args.get('model')
logger.info("file: %s, dataset: %s, model: %s", filename, dataset, model)
predictions = predict([filename],
"mapillary",
"semantic_segmentation",
aggregate=True,
output_dir=PREDICT_FOLDER)
return jsonify(predictions)
def prediction():
"""Route to a jsonified version of deep learning model predictions, for
client tool
Returns
-------
dict
Deep learning model predictions
"""
filename = os.path.basename(request.args.get("img"))
filename = os.path.join(app.config["UPLOAD_FOLDER"], filename)
dataset = request.args.get("dataset")
model = request.args.get("model")
logger.info("file: %s, dataset: %s, model: %s", filename, dataset, model)
predictions = predict(
[filename],
"mapillary",
"semseg",
output_dir=PREDICT_FOLDER,
)
return jsonify(predictions)
