def route(options, body):
#options: a list of strings that defines what functionality the user requested, includes user passed variables
routes = {
'account': account,
'data': data,
'social': social,
'picture': picture,
'ml': ml
}
try:
main_func = routes[
options[0]] #something in the routes dictionary above
assert hasattr(main_func, options[1])
func_call = getattr(main_func,
options[1]) #something like account.login
#TODO assert that everything checks out
#actually call the thing
except:
return constants.respond(err="Bad Path", statusCode="400")
if (body is None):
func_call(options)
else:
func_call(body)
def upload_annotations(options):
print(options)
#TODO add support for CSV
#TODO assert all arguments are present
#Take the annotations (a dictionary)
bucket_name = "machine-learning2019"
print("ANNOTA", options['annotations'])
annotations = options['annotations']
s3 = boto3.resource("s3")
bucket_source={"Bucket":bucket_name, "Key":"/"}
for pic_url in annotations:
print("PIC_URL", pic_url)
bytesIO = BytesIO()
print("type pic-rul annot", annotations[pic_url], type(annotations[pic_url]))
bytesIO.write(json.dumps(annotations[pic_url]).encode())
bytesIO.seek(0)
#TODO get model name and input that instead of default resnet
s3.meta.client.upload_fileobj(bytesIO, bucket_name,
'resnet_keras.h5.dir/annotated_images/annotations/{}.json'.format(pic_url.split('.')[0]))
src_unann = 'resnet_keras.h5.dir/unannotated_images/{}'.format(pic_url)
dest_ann = 'resnet_keras.h5.dir/annotated_images/{}'.format(pic_url)
bucket_source["Key"]=src_unann
print("SRVC_UNANN", src_unann, bucket_source["Key"])
#s3.meta.client.download("resnet_keras.h5.dir/unannotated_images/1247343586733.jpeg", bucket_name, ".")
# Copy object A as object B
s3.Object(bucket_name, dest_ann).copy_from(
CopySource=bucket_source)
# Delete the former object A
s3.Object(bucket_name, src_unann).delete()
#s3.meta.client.copy(CopySource=bucket_source, Bucket=bucket_name, Key=dest_ann)
return constants.respond(statusCode="200")
def register(options):
#options: a list of variables
given_email = options['email']
given_password = options['pass']
#HASH here
given_password = auth.hash_password(given_password)
#First connect to the table
client = boto3.client('rds-data')
#Check if the user already exist in the database and throw a 400 statuscode if they do
existing_user = client.execute_statement(
secretArn=constants.SECRET_ARN,
database=constants.DB_NAME,
resourceArn=constants.ARN,
sql="SELECT email FROM UserData WHERE email = '%s'" % (given_email))
if (existing_user['records'] != []):
print("user exists already")
constants.ERR = "User already exists"
constants.STATUS_CODE = 409
return
existing_user = client.execute_statement(
secretArn=constants.SECRET_ARN,
database=constants.DB_NAME,
resourceArn=constants.ARN,
sql=
"INSERT INTO UserData (email, pass, type, name) VALUES ('%s','%s','%s','%s')"
% (given_email, given_password, 0, given_email))
#Return success
return constants.respond(statusCode="200") #OK
def mobile_classify(options):
print("MOBIL CLASS")
pic_source = base64.b64decode(options['pic'])
model=str(options["model"])
#create a client to make requests to ml api
http = urllib3.PoolManager()
bytesIO = BytesIO()
bytesIO.write(pic_source)
bytesIO.seek(0)
s3 = boto3.resource("s3")
bucket_name = "machine-learning2019"
now = str(time.monotonic_ns())
pic_path = "{}.dir/unannotated_images/{}.jpeg".format(model, now)
#upload the file that was passed to the S3 bucket
s3 = boto3.resource("s3")
bucket_name = "machine-learning2019"
s3.meta.client.upload_fileobj(bytesIO, bucket_name, pic_path)
#s3.meta.client.upload_file(pic_source, bucket_name, pic_path)
print("MOBILE CLASS UPLAODED IMAGE TO S3")
#TODO replace with actual request
r = http.request('GET', 'ec2-3-18-109-238.us-east-2.compute.amazonaws.com:3000/predict?modelName={}&imageURL={}'.format(model, pic_path))
#resp stores the classifier prediction
resp = json.loads(r.data.decode('utf-8'))
return constants.respond(statusCode="200", res=resp)
def classify(options):
#options:
# picture:(str) a picture to classify. path to file
# model: (str) the model to classify with
#TODO assert these are proper types
pic_source = str(options["pic"])
model=str(options["model"])
#create a client to make requests to ml api
http = urllib3.PoolManager()
bytesIO = BytesIO()
bytesIO.write(http.request("GET", pic_source).data)
bytesIO.seek(0)
s3 = boto3.resource("s3")
bucket_name = "machine-learning2019"
now = str(time.monotonic_ns())
pic_path = "{}.dir/unannotated_images/{}.jpeg".format(model, now)
#upload the file that was passed to the S3 bucket
s3 = boto3.resource("s3")
bucket_name = "machine-learning2019"
s3.meta.client.upload_fileobj(bytesIO, bucket_name, pic_path)
#s3.meta.client.upload_file(pic_source, bucket_name, pic_path)
#TODO replace with actual request
r = http.request('GET', 'ec2-3-18-109-238.us-east-2.compute.amazonaws.com:3000/predict?modelName={}&imageURL={}'.format(model, pic_path))
#resp stores the classifier prediction
resp = json.loads(r.data.decode('utf-8'))
return constants.respond(statusCode="200", res=resp)
def upload_model(options):
model_url = options["modelURL"]
model_name = options['modelName']
model_type = options['modelType']
#TODO make call to ML API
http = urllib3.PoolManager()
req_url = 'http://ec2-3-18-109-238.us-east-2.compute.amazonaws.com:3000/upload/new?modelFileURL={}&modelName={}&modelType={}'.format(model_url, model_name, model_type)
r = http.request('POST', req_url)
print(r.data.decode("utf-8"))
return constants.respond(statusCode="200", res = json.loads(r.data.decode('utf-8')))
def main(event, context):
constants.ERR = None
constants.STATUS_CODE = "200"
constants.RES = "Success"
path = event['path']
httpType = event['httpMethod']
path = parse_path(path)
# if('' in path):
# print("lambda just tried to reinvoke itself. terminating")
# return constants.respond(err="Parse error, lambda reinvoked", statusCode = "403")
print(path)
if (httpType == "GET"):
print("Routing GET")
route(path, None)
else:
print("Routing not GET")
try:
body = event['body']
except:
return constants.respond(err="NO BODY", statusCode="404")
try:
body = json.loads(body)
except ValueError as e:
print("body error")
return constants.respond(err="BODY ERROR, CHECK FORMAT",
statusCode="400")
route(path, body)
return constants.respond(constants.ERR, constants.RES,
constants.STATUS_CODE)
def list(options):
http = urllib3.PoolManager()
r = http.request('GET', 'http://ec2-3-18-109-238.us-east-2.compute.amazonaws.com:3000/models/list')
print("YOOOOOO", r.data)
resp = json.loads(r.data.decode('utf-8'))
return constants.respond(statusCode="200", res=resp)
def dl_unannotated_imgs(options):
return constants.respond(statusCode="200", res={'img1':"23879ryghfkvh2039toirghir9u3289tuoirshnksajshkvsjgh2938ut98urfhvjnkjhow8"})
