deepdetect_ruby 1.0.1

data/docs/Finetuning.md

@@ -0,0 +1,241 @@
+# Finetuning
+```
+DeepdetectRuby::Service.create({name: "selfie_finetuning", model_name: "selfie_finetuning", finetuning: true, weights: "model_iter_16368.caffemodel"})
+
+DeepdetectRuby::Service.create({name: "selfie_finetuning_t1", finetuning: true, weights: "model_iter_16368.caffemodel"})
+
+DeepdetectRuby::Service.create({name: "selfie_finetuning_t2", finetuning: true, weights: "model_iter_16368.caffemodel"})
+
+DeepdetectRuby::Service.delete("selfie_finetuning")
+
+DeepdetectRuby::Train.get_status(service: "selfie_finetuning")
+
+DeepdetectRuby::Train.get_status(service: "selfie_finetuning_t1")
+
+DeepdetectRuby::Service.finetuning({name: "selfie_finetuning", finetuning: true, weights: "model_iter_16368.caffemodel", repository: "/home/ubuntu/training/finetuning/selfie", gpu: true, iterations: 16368, test_interval: 500})
+
+DeepdetectRuby::Service.finetuning({name: "person_yes_no_f1", finetuning: true, weights: "model_iter_5213.caffemodel", repository: "/home/ubuntu/training/finetuning/person_yes_no", gpu: true, iterations: 10000, test_interval: 500})
+
+DeepdetectRuby::Service.finetuning({name: "selfie_f1", finetuning: true, weights: "model_iter_16368.caffemodel", repository: "/home/ubuntu/training/finetuning/selfie", gpu: true, iterations: 18000, test_interval: 500})
+
+DeepdetectRuby::Service.finetuning({name: "selfie_f2", finetuning: true, weights: "model_iter_18000.caffemodel", repository: "/home/ubuntu/training/finetuning/selfie_f2", gpu: true, iterations: 18000, test_interval: 500})
+
+DeepdetectRuby::Service.finetuning({name: "more_than_one_f1", finetuning: true, weights: "model_iter_15000.caffemodel", repository: "/home/ubuntu/training/finetuning/more_than_one", gpu: true, iterations: 18000, test_interval: 500})
+
+DeepdetectRuby::Service.finetuning({name: "selfie_f2_1", finetuning: true, weights: "model_iter_18000.caffemodel", repository: "/home/ubuntu/training/finetuning/selfie_f2", gpu: true, iterations: 18000, test_interval: 500})
+
+DeepdetectRuby::Service.finetuning({name: "selfie_f3", finetuning: true, weights: "model_iter_16368.caffemodel", repository: "/home/ubuntu/training/finetuning/selfie_f2", gpu: true, iterations: 10230, test_interval: 500, measure_index: 2})
+
+DeepdetectRuby::Service.finetuning({name: "selfie_f4", finetuning: true, weights: "model_iter_16368.caffemodel", repository: "/home/ubuntu/training/finetuning/selfie_f4", gpu: true, iterations: 20002, test_interval: 500, measure_index: 1, batch_size: 32, test_batch_size: 32})
+
+```
+## Training new updated models with finetuning
+```
+DeepdetectRuby::Train.launch({service: "selfie_finetuning", repository: "/home/ubuntu/training/finetuning/selfie", gpu: true, iterations: 16368})
+
+DeepdetectRuby::Train.launch({service: "selfie_finetuning_t1", repository: "/home/ubuntu/training/finetuning/selfie", gpu: true, iterations: 16368, test_interval: 500})
+
+DeepdetectRuby::Train.launch({service: "selfie_finetuning_t2", repository: "/home/ubuntu/training/finetuning/selfie", gpu: true, iterations: 16368, test_interval: 500, finetuning: true})
+```
+--------------------------------------------------------------
+Create a service
+curl -X PUT "http://localhost:8080/services/selfie" -d "{\"mllib\":\"caffe\",\"description\":\"selfie classification\",\"type\":\"supervised\",\"parameters\":{\"input\":{\"connector\":\"image\",\"height\":224,\"width\":224},\"mllib\":{\"nclasses\":304}},\"model\":{\"repository\":\"/home/ubuntu/projects/deepdetect_u1/models/selfie_f40\"}}"
+--------------------------------------------------------------
+Finetuning model selfie_f3
+curl -X PUT "http://localhost:8080/services/selfie" -d "{\"mllib\":\"caffe\",\"description\":\"selfie classification\",\"type\":\"supervised\",\"parameters\":{\"input\":{\"connector\":\"image\",\"height\":224,\"width\":224},\"mllib\":{\"nclasses\":2, \"mirror\":true, \"weights\":\"model_iter_16368.caffemodel\",\"finetuning\":true, \"gpu\":true}},\"model\":{\"repository\":\"/home/ubuntu/projects/deepdetect_u1/models/selfie_f40\"}}"
+
+curl -X POST "http://localhost:8080/train" -d "{\"service\":\"selfie\",\"async\":true,\"parameters\":{\"mllib\":{\"gpu\":true,\"net\":{\"batch_size\":32},\"solver\":{\"test_interval\":500,\"iterations\":18000,\"base_lr\":0.0001,\"stepsize\":1000,\"gamma\":0.9}},\"input\":{\"connector\":\"image\",\"test_split\":0.1,\"shuffle\":true,\"width\":224,\"height\":224},\"output\":{\"measure\":[\"acc-5\",\"mcll\",\"f1\"]}},\"data\":[\"/home/ubuntu/training/finetuning/selfie_f4\"]}"
+--------------------------------------------------------------
+
+curl -X POST "http://localhost:8080/train" -d "{\"service\":\"selfie_finetuning\",\"async\":true,\"parameters\":{\"mllib\":{\"gpu\":true,\"net\":{\"batch_size\":32},\"solver\":{\"test_interval\":500,\"iterations\":10000,\"base_lr\":0.001,\"stepsize\":1000,\"gamma\":0.9}},\"input\":{\"connector\":\"image\",\"test_split\":0.1,\"shuffle\":true,\"width\":224,\"height\":224},\"output\":{\"measure\":[\"acc\",\"mcll\",\"f1\"]}},\"data\":[\"/home/ubuntu/training/finetuning/selfie\"]}"
+
+--------------------------
+DeepdetectRuby::Predict.predict({service: "selfie_finetuning", image_url: "http://cdn.feels.com/instagram/asos/1246490803437553418_1557663056.jpg"})
+
+DeepdetectRuby::Predict.predict({service: "person_yes_no_f1", image_url: "http://cdn.feels.com/instagram/asos/1246490803437553418_1557663056.jpg"})
+## Before finetuning
+:predictions => {
+                :uri => "http://cdn.feels.com/instagram/asos/1246490803437553418_1557663056.jpg",
+            :classes => [
+                [0] {
+                    :prob => 0.5820828080177307,
+                     :cat => "selfie"
+                },
+                [1] {
+                    :last => true,
+                    :prob => 0.4179171919822693,
+                     :cat => "no_selfie"
+                }
+            ]
+        }
+
+## After finetuning
+:predictions => {
+                :uri => "http://cdn.feels.com/instagram/asos/1246490803437553418_1557663056.jpg",
+            :classes => [
+                [0] {
+                    :prob => 0.9999997615814209,
+                     :cat => "selfie"
+                },
+                [1] {
+                    :last => true,
+                    :prob => 2.4337700210708135e-07,
+                     :cat => "no_selfie"
+                }
+            ]
+        }
+
+-----------------------------
+http://feels.imgix.net/instagram/asos/1252542985950992360_2206765260.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top
+Origin:
+https://www.instagram.com/p/BFh7TWfhBfo/
+{"normal_resolution":0.990632,"bad_resolution":0.00936781,"normal_lighting":0.987771,"poor_lighting":0.0122286,"someone":0.901468,"no_one":0.0985324,"less_than_one":0.998956,"more_than_one":0.0010435,"men":0.562763,"women":0.437237,"age_above_30":0.40452,"selfie":0.880388,"no_selfie":0.119612,"no_mirrorshot":0.962171,"mirrorshot":0.0378293}
+
+After finetuning:
+---> DeepdetectRuby::Predict.predict({service: "selfie_finetuning_t1", image_url: "http://feels.imgix.net/instagram/asos/1252542985950992360_2206765260.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top"})
+
+:predictions => {
+                :uri => "http://feels.imgix.net/instagram/asos/1252542985950992360_2206765260.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top",
+            :classes => [
+                [0] {
+                    :prob => 0.9978917241096497,
+                     :cat => "no_selfie"
+                },
+                [1] {
+                    :last => true,
+                    :prob => 0.002108336426317692,
+                     :cat => "selfie"
+                }
+            ]
+        }
+
+-----------------------------
+http://feels.imgix.net/instagram/asos/1252353525430668710_1814295050.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top
+Origin: https://www.instagram.com/p/BFhQOVogM2m/
+{"normal_resolution":0.825466,"bad_resolution":0.174534,"normal_lighting":0.884736,"poor_lighting":0.115264,"someone":0.999343,"no_one":0.000656711,"more_than_one":0.845909,"less_than_one":0.154091,"men":1,"women":1.60751e-10,"age_under_16":0.917247,"selfie":0.976695,"no_selfie":0.0233054,"no_mirrorshot":0.996792,"mirrorshot":0.00320824}
+
+After finetuning:
+---> DeepdetectRuby::Predict.predict({service: "selfie_finetuning_t1", image_url: "http://feels.imgix.net/instagram/asos/1252353525430668710_1814295050.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top"})
+------------------------------
+http://feels.imgix.net/instagram/asos/1251924354355760960_14740578.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top
+{"normal_resolution":0.996735,"bad_resolution":0.00326459,"normal_lighting":0.971896,"poor_lighting":0.0281036,"someone":0.878907,"no_one":0.121093,"less_than_one":0.935041,"more_than_one":0.0649591,"women":0.999993,"men":0.00000736875,"age_above_30":0.567047,"selfie":0.863574,"no_selfie":0.136426,"no_mirrorshot":0.996853,"mirrorshot":0.00314677}
+DeepdetectRuby::Predict.predict({service: "selfie_finetuning_t1", image_url: "http://feels.imgix.net/instagram/asos/1251924354355760960_14740578.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top"})
+:predictions => {
+                :uri => "http://feels.imgix.net/instagram/asos/1251924354355760960_14740578.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top",
+            :classes => [
+                [0] {
+                    :prob => 1.0,
+                     :cat => "selfie"
+                },
+                [1] {
+                    :last => true,
+                    :prob => 5.714219852848146e-08,
+                     :cat => "no_selfie"
+                }
+            ]
+        }
+-------------------------------------
+http://feels.imgix.net/instagram/asos/1251985366878743528_10194825.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top
+https://www.instagram.com/p/BFf8g7Qy2vo/
+{"someone":1,"no_one":0,"selfie":1,"women":1,"men":0,"no_selfie":0,"normal_resolution":0.999719,"bad_resolution":0.000280655,"normal_lighting":0.998396,"poor_lighting":0.00160362,"less_than_one":0.912216,"more_than_one":0.0877836,"age_above_30":0.469602}
+
+DeepdetectRuby::Predict.predict({service: "selfie_finetuning_t1", image_url: "http://feels.imgix.net/instagram/asos/1252033944498407117_1223425960.jpg?w=530&h=530&dpr=2&fit=crop&crop=faces,top"})
+
+DeepdetectRuby::Predict.predict({service: "selfie_finetuning_t2", image_url: "http://feels.imgix.net/instagram/asos/1252404500268043866_27876266.jpg"})
+
+DeepdetectRuby::Predict.predict({service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255170406470979359_3253319674.jpg"})
+
+DeepdetectRuby::Predict.predict({service: "selfie_f1", image_url: "//feels.imgix.net/instagram/asos/1255355896542752718_27057010.jpg"})
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255266658690557670_11397328.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255271338485499066_775122576.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255424592408535294_185254919.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255390340978924420_15115085.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255662739004227204_330560670.jpg") --> not ok, must be selfie
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255602514657391462_15783947.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255622208398307634_317677827.jpg") --> not ok, must be selfie
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255534513632110008_553559664.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255455561021981994_197430064.jpg") --> not ok, must be selfie
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255399707491145944_1457907277.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255337649249306267_558272928.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://feels.imgix.net/instagram/asos/1255237247935534338_238585596.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://cdn.feels.com/instagram/asos/1256146576133469486_222484236.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f1", image_url: "http://cdn.feels.com/instagram/asos/1256146220572730237_30723063.jpg")
+
+DeepdetectRuby::Predict.predict(service: "more_than_one_f1", image_url: "http://cdn.feels.com/instagram/asos/1256909233581381035_264705436.jpg")
+
+DeepdetectRuby::Predict.predict(service: "more_than_one_f1", image_url: "http://cdn.feels.com/instagram/asos/1256902797330248703_3013321051.jpg")
+
+DeepdetectRuby::Predict.predict(service: "more_than_one_f1", image_url: "http://feels.imgix.net/instagram/asos/1256951961487542817_1462068201.jpg")
+
+DeepdetectRuby::Predict.predict(service: "selfie_f2_1", image_url: "http://feels.imgix.net/instagram/newlookfashion/1231213665197843799_1452988672.jpg") --> error
+
+DeepdetectRuby::Predict.predict(service: "selfie_f3", image_url: "http://feels.imgix.net/instagram/newlookfashion/1231213665197843799_1452988672.jpg")
+
+curl -X POST "http://localhost:8080/predict" -d "{\"service\":\"selfie_f3\",\"parameters\":{\"output\":{\"best\":2}},\"data\":[\"http://feels.imgix.net/instagram/newlookfashion/1231213665197843799_1452988672.jpg\"]}"
+--------------------------
+http://cdn.feels.com/instagram/newlookfashion/1221769581631587626_433120464.jpg
+old predict: --> bad, wrong result
+DeepdetectRuby::Predict.predict(service: "selfie", image_url: "http://cdn.feels.com/instagram/newlookfashion/1221769581631587626_433120464.jpg")
+:predictions => {
+                :uri => "http://cdn.feels.com/instagram/newlookfashion/1221769581631587626_433120464.jpg",
+            :classes => [
+                [0] {
+                    :prob => 0.95999675989151,
+                     :cat => "selfie"
+                },
+                [1] {
+                    :last => true,
+                    :prob => 0.040003255009651184,
+                     :cat => "no_selfie"
+                }
+            ]
+        }
+
+new predict: --> better, more accuracy
+DeepdetectRuby::Predict.predict(service: "selfie_f3", image_url: "http://cdn.feels.com/instagram/newlookfashion/1221769581631587626_433120464.jpg")
+:predictions => {
+      :uri => "https://feels.imgix.net/instagram/newlookfashion/1221769581631587626_433120464.jpg",
+  :classes => [
+      [0] {
+          :prob => 0.8537927865982056,
+           :cat => "no_selfie"
+      },
+      [1] {
+          :last => true,
+          :prob => 0.14620719850063324,
+           :cat => "selfie"
+      }
+  ]
+}
+
+-------------------
+OpenCV Error: Assertion failed (buf.data && buf.isContinuous()) in imdecode_, file /build/buildd/opencv-2.4.8+dfsg1/modules/highgui/src/loadsave.cpp, line 307
+
+ERROR - 02:32:08 - service selfie_f3 mllib bad param: /build/buildd/opencv-2.4.8+dfsg1/modules/highgui/src/loadsave.cpp:307: error: (-215) buf.data && buf.isContinuous() in function imdecode_
+
+------------------------
+DeepdetectRuby::Predict.predict(service: "mirrorshot_f1", image_url: "http://cdn.feels.com/instagram/converse/1286789533837253555_412988956.jpg")
+DeepdetectRuby::Predict.predict(service: "mirrorshot_f1", image_url: "http://cdn.feels.com/instagram/converse/1286744821742913513_1637327762.jpg")
+
+------------------------
+DeepdetectRuby::Predict.predict(service: "more_than_one_c", image_url: "http://feels.imgix.net/instagram/asos/1256951961487542817_1462068201.jpg")
+
+DeepdetectRuby::Predict.predict(service: "more_than_one_c", image_url: "http://cdn.feels.com/instagram/missguided/1293683001124816735_215780245.jpg")
+
+
+-------------------
+## Rsync

data/docs/convolutional_neural_networks.py

@@ -0,0 +1,28 @@
+from dd_client import DD
+
+model_repo = '/home/me/models/sent_en_char'
+nclasses = 2
+
+# setting up DD client
+host = '127.0.0.1'
+sname = 'sent_en'
+description = 'English sentiment classification'
+mllib = 'caffe'
+dd = DD(host)
+dd.set_return_format(dd.RETURN_PYTHON)
+
+# creating ML service
+model = {'repository':model_repo}
+parameters_input = {'connector':'txt','characters':True,'sequence':140,'alphabet':"abcdefghijklmnopqrstuvwxyz0123456789,;.!?'"}
+parameters_mllib = {'nclasses':nclasses}
+parameters_output = {}
+dd.put_service(sname,model,description,mllib,
+               parameters_input,parameters_mllib,parameters_output)
+
+# classifying a single piece of text:
+parameters_input = {}
+parameters_mllib = {}
+parameters_output = {}
+data = ['Chilling in the West Indies']
+classif = dd.post_predict(sname,data,parameters_input,parameters_mllib,parameters_output)
+print classif

data/docs/dd_client.py

@@ -0,0 +1,379 @@
+"""
+DeepDetect Python client
+
+Licence:
+Copyright (c) 2015 Emmanuel Benazera
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+"""
+
+try:
+    import urllib.request as urllib2
+except ImportError:
+    import urllib2
+
+try:
+    import http.client as httplib
+except ImportError:
+    import httplib
+
+
+import os.path
+import json
+import uuid
+import datetime
+
+
+VERBOSE=False
+DD_TIMEOUT = 2000 # seconds, for long blocking training alls, as needed
+
+def LOG(msg):
+    """Output a log message."""
+    # XXX: may want to use python log manager classes instead of this stupid print
+    if VERBOSE:
+        msg = str(datetime.datetime.now()) + ' ' + msg
+        print (msg)
+
+### Exception classes :
+
+class DDCommunicationError(Exception):
+    def __init__(self, url, http_method,  headers, body, response=None):
+        self.msg = """DeepDetect Communication Error"""
+        self.http_method = http_method
+        self.req_headers = headers
+        self.req_body = body
+        self.url = url
+        self.res_headers = None
+        if response is not None:
+            self.res_headers = response.get_info()
+
+    def __str__(self):
+        msg = "%s %s\n"%(str(self.http_method),str(self.url))
+        for h,v in self.req_headers.iteritems():
+            msg += "%s:%s\n"%(h,v)
+        msg += "\n"
+        if self.req_body is not None:
+            msg += str(self.req_body)[:100]
+        msg += "\n"
+        msg += "--\n"
+        msg += str(self.res_headers)
+        msg += "\n"
+        return msg
+
+class DDDataError(Exception):
+    def __init__(self, url, http_method,  headers, body, data=None):
+        self.msg = "DeepDetect Data Error"
+        self.http_method = http_method
+        self.req_headers = headers
+        self.req_body = body
+        self.url = url
+        self.data = data
+
+    def __str__(self):
+        msg = "%s %s\n"%(str(self.http_method),str(self.url))
+        if self.data is not None:
+            msg += str(self.data)[:100]
+        msg += "\n"
+        for h,v in self.req_headers.iteritems():
+            msg += "%s:%s\n"%(h,v)
+        msg += "\n"
+        if self.req_body is not None:
+            msg += str(self.req_body)
+        msg += "\n"
+        msg += "--\n"
+        msg += str(self.data)
+        msg += "\n"
+        return msg
+
+API_METHODS_URL = {
+    "0.1" : {
+        "info":"/info",
+        "services":"/services",
+        "train":"/train",
+        "predict":"/predict"
+    }
+}
+
+class DD(object):
+    """HTTP requests to the DeepDetect server
+
+    """
+
+    # return types
+    RETURN_PYTHON=0
+    RETURN_JSON=1
+    RETURN_NONE=2
+
+    __HTTP=0
+    __HTTPS=1
+
+    def __init__(self,host="localhost",port=8080,proto=0,apiversion="0.1"):
+        """ DD class constructor
+        Parameters:
+        host -- the DeepDetect server host
+        port -- the DeepDetect server port
+        proto -- user http (0,default) or https connection
+        """
+        self.apiversion = apiversion
+        self.__urls = API_METHODS_URL[apiversion]
+        self.__host = host
+        self.__port = port
+        self.__proto = proto
+        self.__returntype=self.RETURN_PYTHON
+        if proto == self.__HTTP:
+            self.__ddurl='http://%s:%d'%(host,port)
+        else:
+            self.__ddurl='https://%s:%d'%(host,port)
+
+
+    def set_return_format(self,f):
+        assert f == self.RETURN_PYTHON or f == self.RETURN_JSON or f == self.RETURN_NONE
+        self.__returntype = f
+
+    def __return_format(self,js):
+        if self.__returntype == self.RETURN_PYTHON:
+            return json.loads(js.decode('utf-8'))
+        elif self.__returntype == self.RETURN_JSON:
+            return js
+        else:
+            return None
+
+    def get(self,method,args=None):
+        """ GET to DeepDetect server """
+        u = self.__ddurl
+        u += method
+        headers = {}
+        if args is not None:
+            sep = "?"
+            for arg,argv in args.iteritems():
+                u += sep
+                sep = "&"
+                u += urllib2.quote(arg)
+                u += '='
+                if argv is not None:
+                    u += urllib2.quote(argv)
+
+        LOG("GET %s"%u)
+        response = None
+        try:
+            req = urllib2.Request(u)
+            response = urllib2.urlopen(req, timeout=DD_TIMEOUT)
+            jsonresponse=response.read()
+        except:
+            raise DDCommunicationError(u,"GET",headers,None,response)
+        LOG(jsonresponse)
+        try:
+            return self.__return_format(jsonresponse)
+        except:
+            raise DDDataError(u,"GET",headers,None,jsonresponse)
+
+    def put(self, method, body):
+        """PUT request to DeepDetect server"""
+
+        LOG("PUT %s\n%s"%(method,body))
+        r = None
+        u = ""
+        headers = {}
+        try:
+            u = self.__ddurl + method
+            if self.__proto == self.__HTTP:
+            #    u = "http://%s:%s%s"%(self.__host,self.__port,method)
+                c=httplib.HTTPConnection(self.__host,self.__port, timeout=DD_TIMEOUT)
+            else:
+            #    u = "https://%s:%s%s"%(self.__host,self.__port,method)
+                c=httplib.HTTPSConnection(self.__host,self.__port, timeout=DD_TIMEOUT)
+            c.request('PUT',method,body,headers)
+            r = c.getresponse()
+            data = r.read()
+        except:
+            raise DDCommunicationError(u,"PUT",headers,body,r)
+        LOG(data)
+        try:
+            return self.__return_format(data)
+        except:
+            raise DDDataError(u,"PUT",headers,body,data)
+
+    def post(self,method,body):
+        """POST request to DeepDetect server"""
+
+        r = None
+        u = ""
+        headers = {}
+        try:
+            u = self.__ddurl + method
+            if self.__proto == self.__HTTP:
+                LOG("curl -X POST 'http://%s:%s%s' -d '%s'"%(self.__host,
+                                                             self.__port,
+                                                             method,
+                                                             body))
+                c=httplib.HTTPConnection(self.__host,self.__port,timeout=DD_TIMEOUT)
+            else:
+                LOG("curl -k -X POST 'https://%s:%s%s' -d '%s'"%(self.__host,
+                                                                 self.__port,
+                                                                 method,
+                                                                 body))
+                c=httplib.HTTPSConnection(self.__host,self.__port, timeout=DD_TIMEOUT)
+            c.request('POST',method,body,headers)
+            r = c.getresponse()
+            data = r.read()
+
+        except:
+            raise DDCommunicationError(u,"POST",headers,body,r)
+
+        # LOG(data)
+        try:
+            return self.__return_format(data)
+        except:
+            import traceback
+            print (traceback.format_exc())
+
+            raise DDDataError(u,"POST",headers,body,data)
+
+    def delete(self, method):
+        """DELETE request to DeepDetect server"""
+
+        LOG("DELETE %s"%(method))
+        r = None
+        u = ""
+        body = ""
+        headers = {}
+        try:
+            u = self.__ddurl + method
+            if self.__proto == self.__HTTP:
+                c=httplib.HTTPConnection(self.__host,self.__port, timeout=DD_TIMEOUT)
+            else:
+                c=httplib.HTTPSConnection(self.__host,self.__port, timeout=DD_TIMEOUT)
+            c.request('DELETE',method,body,headers)
+            r = c.getresponse()
+            data = r.read()
+        except:
+            raise DDCommunicationError(u,"DELETE",headers,None,r)
+
+        LOG(data)
+        try:
+            return self.__return_format(data)
+        except:
+            raise DDDataError(u,"DELETE",headers,None,data)
+
+
+    # API methods
+
+    def info(self):
+        """Info on the DeepDetect server"""
+        return self.get(self.__urls["info"])
+
+
+    # - PUT services
+    # - GET services
+    # - DELETE services
+    def put_service(self,sname,model,description,mllib,parameters_input,parameters_mllib,parameters_output,mltype='supervised'):
+        """
+        Create a service
+        Parameters:
+        sname -- service name as a resource
+        model -- dict with model location and optional templates
+        description -- string describing the service
+        mllib -- ML library name, e.g. caffe
+        parameters_input -- dict of input parameters
+        parameters_mllib -- dict ML library parameters
+        parameters_output -- dict of output parameters
+        """
+        body={"description":description,"mllib":mllib,"type":mltype,
+              "parameters":{"input":parameters_input,"mllib":parameters_mllib,"output":parameters_output},
+              "model":model}
+        return self.put(self.__urls["services"] + '/%s'%sname,json.dumps(body))
+
+    def get_service(self,sname):
+        """
+        Get information about a service
+        Parameters:
+        sname -- service name as a resource
+        """
+        return self.get(self.__urls["services"] + '/%s'%sname)
+
+    def delete_service(self,sname,clear=None):
+        """
+        Delete a service
+        Parameters:
+        sname -- service name as a resource
+        clear -- 'full','lib' or 'mem', optionally clears model repository data
+        """
+        qs = self.__urls["services"] + '/%s'%sname
+        if clear:
+            qs += '?clear=' + clear
+        return self.delete(qs)
+
+
+    # PUT/POST /train
+    # GET /train
+    # DELETE /train
+
+    def post_train(self,sname,data,parameters_input,parameters_mllib,parameters_output,async=True):
+        """
+        Creates a training job
+        Parameters:
+        sname -- service name as a resource
+        async -- whether to run the job as non-blocking
+        data -- array of input data / dataset for training
+        parameters_input -- dict of input parameters
+        parameters_mllib -- dict ML library parameters
+        parameters_output -- dict of output parameters
+        """
+        body={"service":sname,"async":async,
+              "parameters":{"input":parameters_input,"mllib":parameters_mllib,"output":parameters_output},
+              "data":data}
+        return self.post(self.__urls["train"],json.dumps(body))
+
+    def get_train(self,sname,job=1,timeout=0,measure_hist=False):
+        """
+        Get information on a non-blocking training job
+        Parameters:
+        sname -- service name as a resource
+        job -- job number on the service
+        timeout -- timeout before obtaining the job status
+        measure_hist -- whether to return the full measure history (e.g. for plotting)
+        """
+        qs=self.__urls["train"] + "?service=" + sname + "&job=" + str(job) + "&timeout=" + str(timeout)
+        if measure_hist:
+            qs += "&parameters.output.measure_hist=true"
+        return self.get(qs)
+
+    def delete_train(self,sname,job=1):
+        """
+        Kills a non-blocking training job
+        Parameters:
+        sname -- service name as a resource
+        job -- job number on the service
+        """
+        qs=self.__urls["train"] + "?service=" + sname + "&job=" + str(job)
+        return self.delete(qs)
+
+
+    # POST /predict
+
+    def post_predict(self,sname,data,parameters_input,parameters_mllib,parameters_output):
+        """
+        Makes prediction from data and model
+        Parameters:
+        sname -- service name as a resource
+        data -- array of data URI to predict from
+        parameters_input -- dict of input parameters
+        parameters_mllib -- dict ML library parameters
+        parameters_output -- dict of output parameters
+        """
+        body={"service":sname,
+              "parameters":{"input":parameters_input,"mllib":parameters_mllib,"output":parameters_output},
+              "data":data}
+        return self.post(self.__urls["predict"],json.dumps(body))
+
+# test
+if __name__ == '__main__':
+    dd = DD()
+    dd.set_return_format(dd.RETURN_PYTHON)
+    inf = dd.info()
+    print (inf)