def preprocess(self, inp):
"""
Default preprocessing method for is to convert the picture to black and white and resize to be 48x48
"""
im = preprocessing_utils.decode_base64_to_image(inp)
im = im.convert('RGB')
im = preprocessing_utils.resize_and_crop(im, (self.image_width, self.image_height))
return np.array(im)
def rebuild(self, dir, data):
"""
Default rebuild method to decode a base64 image
"""
im = preprocessing_utils.decode_base64_to_image(data)
timestamp = datetime.datetime.now()
filename = f'input_{timestamp.strftime("%Y-%m-%d-%H-%M-%S")}.png'
im.save('{}/{}'.format(dir, filename), 'PNG')
return filename
def rebuild_flagged(self, dir, msg):
"""
Default rebuild method to decode a base64 image
"""
im = preprocessing_utils.decode_base64_to_image(msg)
timestamp = time.time() * 1000
filename = f'input_{timestamp}.png'
im.save(f'{dir}/{filename}', 'PNG')
return filename
def rebuild_flagged(self, dir, msg):
"""
Default rebuild method to decode a base64 image
"""
im = preprocessing_utils.decode_base64_to_image(msg)
timestamp = datetime.datetime.now()
filename = f'output_{timestamp.strftime("%Y-%m-%d-%H-%M-%S")}.png'
im.save(f'{dir}/{filename}', 'PNG')
return filename
def rebuild_flagged(self, dir, msg):
"""
Default rebuild method to decode a base64 image
"""
im = preprocessing_utils.decode_base64_to_image(msg)
timestamp = datetime.datetime.now()
filename = 'output_{}.png'.format(timestamp.
strftime("%Y-%m-%d-%H-%M-%S"))
im.save('{}/{}'.format(dir, filename), 'PNG')
return filename
def preprocess(self, inp):
"""
Default preprocessing method for is to convert the picture to black and white and resize to be 48x48
"""
im = preprocessing_utils.decode_base64_to_image(inp)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
im = im.convert(self.image_mode)
im = preprocessing_utils.resize_and_crop(im, (self.image_width, self.image_height))
return np.array(im)
def preprocess(self, inp):
"""
Default preprocessing method for the SketchPad is to convert the sketch to black and white and resize 28x28
"""
im_transparent = preprocessing_utils.decode_base64_to_image(inp)
im = PIL.Image.new("RGBA", im_transparent.size, "WHITE") # Create a white background for the alpha channel
im.paste(im_transparent, (0, 0), im_transparent)
im = im.convert('L')
if self.invert_colors:
im = PIL.ImageOps.invert(im)
im = im.resize((self.image_width, self.image_height))
if self.flatten:
array = np.array(im).flatten().reshape(1, self.image_width * self.image_height)
else:
array = np.array(im).flatten().reshape(1, self.image_width, self.image_height)
return array
def preprocess(self, inp):
"""
Default preprocessing method for the SketchPad is to convert the sketch to black and white and resize 28x28
"""
im = preprocessing_utils.decode_base64_to_image(inp)
im = im.convert('L')
if self.invert_colors:
im = ImageOps.invert(im)
im = preprocessing_utils.resize_and_crop(im, (self.image_width, self.image_height))
if self.flatten:
array = np.array(im).flatten().reshape(1, self.image_width * self.image_height)
else:
array = np.array(im).flatten().reshape(1, self.image_width, self.image_height)
array = array * self.scale + self.shift
array = array.astype(self.dtype)
return array
def preprocess(self, inp):
"""
Default preprocessing method for is to convert the picture to black and white and resize to be 48x48
"""
im = preprocessing_utils.decode_base64_to_image(inp)
with warnings.catch_warnings():
warnings.simplefilter("ignore")
im = im.convert(self.image_mode)
im = preprocessing_utils.resize_and_crop(im, (self.image_width, self.image_height))
im = np.array(im).flatten()
im = im * self.scale + self.shift
if self.num_channels is None:
array = im.reshape(1, self.image_width, self.image_height)
else:
array = im.reshape(1, self.image_width, self.image_height, self.num_channels)
return array
def do_POST(self):
# Read body of the request.
if self.path == "/api/predict/":
# Make the prediction.
self._set_headers()
data_string = self.rfile.read(
int(self.headers["Content-Length"]))
msg = json.loads(data_string)
processed_input = interface.input_interface.preprocess(
msg["data"])
prediction = interface.predict(processed_input)
processed_output = interface.output_interface.postprocess(
prediction)
output = {"action": "output", "data": processed_output}
if interface.saliency is not None:
import numpy as np
saliency = interface.saliency(interface.model_obj,
processed_input, prediction)
output['saliency'] = saliency.tolist()
# Prepare return json dictionary.
self.wfile.write(json.dumps(output).encode())
elif self.path == "/api/flag/":
self._set_headers()
data_string = self.rfile.read(
int(self.headers["Content-Length"]))
msg = json.loads(data_string)
flag_dir = os.path.join(directory_to_serve, FLAGGING_DIRECTORY)
os.makedirs(flag_dir, exist_ok=True)
output = {
'input':
interface.input_interface.rebuild_flagged(flag_dir, msg),
'output':
interface.output_interface.rebuild_flagged(flag_dir, msg),
'message':
msg['data']['message']
}
with open(os.path.join(flag_dir, FLAGGING_FILENAME),
'a+') as f:
f.write(json.dumps(output))
f.write("\n")
#TODO(abidlabs): clean this up
elif self.path == "/api/auto/rotation":
from gradio import validation_data, preprocessing_utils
import numpy as np
self._set_headers()
data_string = self.rfile.read(
int(self.headers["Content-Length"]))
msg = json.loads(data_string)
img_orig = preprocessing_utils.decode_base64_to_image(
msg["data"])
img_orig = img_orig.convert('RGB')
img_orig = img_orig.resize((224, 224))
flag_dir = os.path.join(directory_to_serve, FLAGGING_DIRECTORY)
os.makedirs(flag_dir, exist_ok=True)
for deg in range(-180, 180 + 45, 45):
img = img_orig.rotate(deg)
img_array = np.array(img) / 127.5 - 1
prediction = interface.predict(
np.expand_dims(img_array, axis=0))
processed_output = interface.output_interface.postprocess(
prediction)
output = {
'input':
interface.input_interface.save_to_file(flag_dir, img),
'output':
interface.output_interface.rebuild_flagged(
flag_dir, {'data': {
'output': processed_output
}}),
'message':
f'rotation by {deg} degrees'
}
with open(os.path.join(flag_dir, FLAGGING_FILENAME),
'a+') as f:
f.write(json.dumps(output))
f.write("\n")
# Prepare return json dictionary.
self.wfile.write(json.dumps({}).encode())
elif self.path == "/api/auto/lighting":
from gradio import validation_data, preprocessing_utils
import numpy as np
from PIL import ImageEnhance
self._set_headers()
data_string = self.rfile.read(
int(self.headers["Content-Length"]))
msg = json.loads(data_string)
img_orig = preprocessing_utils.decode_base64_to_image(
msg["data"])
img_orig = img_orig.convert('RGB')
img_orig = img_orig.resize((224, 224))
enhancer = ImageEnhance.Brightness(img_orig)
flag_dir = os.path.join(directory_to_serve, FLAGGING_DIRECTORY)
os.makedirs(flag_dir, exist_ok=True)
for i in range(9):
img = enhancer.enhance(i / 4)
img_array = np.array(img) / 127.5 - 1
prediction = interface.predict(
np.expand_dims(img_array, axis=0))
processed_output = interface.output_interface.postprocess(
prediction)
output = {
'input':
interface.input_interface.save_to_file(flag_dir, img),
'output':
interface.output_interface.rebuild_flagged(
flag_dir, {'data': {
'output': processed_output
}}),
'message':
f'brighting adjustment by a factor of {i}'
}
with open(os.path.join(flag_dir, FLAGGING_FILENAME),
'a+') as f:
f.write(json.dumps(output))
f.write("\n")
# Prepare return json dictionary.
self.wfile.write(json.dumps({}).encode())
else:
self.send_error(404, 'Path not found: %s' % self.path)
