Update: Accuracy fomula

cope2n-api/fwd/settings.py

@@ -238,6 +238,13 @@ SUBS = {
     "SEAO": "seao"
     }
 
+FIELDS_BY_SUB = {
+    "SG": {"imei": ["imei_number"],
+            "invoice": ["retailername", "purchase_date"]},
+    "default": {"imei": ["imei_number"],
+            "invoice": ["retailername", "invoice_no", "purchase_date"]},
+    }
+
 BAD_THRESHOLD = 0.75
 NEED_REVIEW = 1.0
 

cope2n-api/fwd_api/celery_worker/process_report_tasks.py

@@ -126,11 +126,13 @@ def create_accuracy_report(report_id, **kwargs):
                                          "invoice_no": mean_list(request_att["acc"]["reviewed"].get("invoice_no", [None]))}
             rq_accuracy = []
 
-            for rpf in _report_files:
-                if sum(len(value_list) for value_list in rpf.reviewed_accuracy.values()):
-                    rq_accuracy += list(chain(*rpf.reviewed_accuracy.values()))
-                elif sum(len(value_list) for value_list in rpf.feedback_accuracy.values()):
-                    rq_accuracy += list(chain(*rpf.feedback_accuracy.values()))
+            for i, _att in enumerate(_atts):
+                if _report_files[i].bad_image_reason in settings.ACC_EXCLUDE_RESEASONS:
+                    continue
+                if sum(len(value_list) for value_list in _att["acc"]["reviewed"].values()):
+                    rq_accuracy += list(chain(*_att["acc"]["reviewed"].values()))
+                elif sum(len(value_list) for value_list in _att["acc"]["feedback"].values()):
+                    rq_accuracy += list(chain(*_att["acc"]["feedback"].values()))
                     
             request.is_required = False
             if len(rq_accuracy) > 0:

cope2n-api/fwd_api/management/commands/migrate-datebase-populate-redemption-id-to-subsidiary.py

@@ -0,0 +1,65 @@
+from django.core.management.base import BaseCommand
+from tqdm import tqdm
+from fwd_api.models import SubscriptionRequestFile, SubscriptionRequest
+from fwd_api.exception.exceptions import InvalidException
+from fwd_api.utils.accuracy import predict_result_to_ready
+import traceback
+import copy
+from django.utils import timezone
+
+KEY = "imei_number"
+VALUE = "None"
+EXPECTED_VALUE = []
+
+class Command(BaseCommand):
+    help = 'Refactor database for image level'
+
+    def add_arguments(self, parser):
+        # Add your command-line arguments here
+        parser.add_argument('start', type=str, help='start date, sample: 2023-01-02T00:00:00+0700')
+        parser.add_argument('end', type=str, help='end date, sample: 2023-01-03T00:00:00+0700')
+
+    def process_request(self, request, result):
+        if len(request.request_id.split(".")[0].split("_")) < 2:
+            return
+        images = SubscriptionRequestFile.objects.filter(request=request)
+        if not request.predict_result:
+            # self.stdout.write(self.style.WARNING(f"Key predict_result not found in {request.request_id}"))
+            return
+        if request.predict_result.get("status", 200) != 200:
+            # self.stdout.write(self.style.WARNING(f"Not a sucess request {request.request_id}"))
+            return
+        
+        if isinstance(request.redemption_id, str) and request.subsidiary is not None:
+            try:
+                request.subsidiary = request.redemption_id[:2]
+                request.save()
+                result['total'] += 1
+                result['subs'].add(request.redemption_id[:2])
+            except Exception as e:
+                print(e)
+                result["failed"] += 1
+
+    def handle(self, *args, **options):
+        start = options['start']
+        end = options['end']
+        result = {'total':0,
+                  'failed':0,
+                  'subs': set()}
+        if start or end:
+            try:
+                start_date = timezone.datetime.strptime(start, '%Y-%m-%dT%H:%M:%S%z') # We care only about day precision only
+                end_date = timezone.datetime.strptime(end, '%Y-%m-%dT%H:%M:%S%z')
+            except Exception as e:
+                print(f"[INFO]: start: {start}")
+                print(f"[INFO]: end: {end}")
+                raise InvalidException(excArgs="Date format")
+            subcription_iter = SubscriptionRequest.objects.filter(created_at__range=(start_date, end_date))
+        else:
+            subcription_iter = SubscriptionRequest.objects.all()
+
+        # file = open("modified.txt", "w")
+        for request in tqdm(subcription_iter.iterator()):
+            self.process_request(request, result)
+        # file.close()
+        self.stdout.write(self.style.SUCCESS('Sample Django management command executed successfully! total/failed: {}/{} - subs: {}'.format(result['total'], result['failed'], result['subs'])))

cope2n-api/fwd_api/utils/accuracy.py

@@ -19,8 +19,6 @@ import json
 from typing import Union, List, Dict
 
 VALID_KEYS = ["retailername", "sold_to_party", "invoice_no", "purchase_date", "imei_number"]
-KEYS_BY_FILE_TYPE = {"imei": ["imei_number"],
-                     "invoice": ["retailername", "invoice_no", "purchase_date"]}
 OPTIONAL_KEYS = ['invoice_no']
 
 class ReportAccumulateByRequest:
@@ -776,7 +774,7 @@ def calculate_accuracy(key_name: str, inference: Dict[str, Union[str, List]], ta
     
     _inference = inference[key_name]
     _target = target[key_name]
-    _will_acc_be_ignored = _acc_will_be_ignored(key_name, _target) 
+    # _will_acc_be_ignored = _acc_will_be_ignored(key_name, _target) 
     _inference, _target = _accuracy_calculate_formatter(_inference, _target)
 
     for i, v in enumerate(_inference):
@@ -793,7 +791,6 @@ def calculate_accuracy(key_name: str, inference: Dict[str, Union[str, List]], ta
                 # "line_acc",
                 # "one_minus_ned_word",
             ])
-        if not _will_acc_be_ignored:
         acc.append(list(score.values())[0])
         data.append([x, y])
     return acc, data
@@ -908,12 +905,11 @@ def calculate_a_request(report, request):
     images = SubscriptionRequestFile.objects.filter(request=request, file_category=FileCategory.Origin.value)
     report_files = []
     for image in images:
-        status, att = calculate_subcription_file(image)
-        atts.append(att)
+        status, att = calculate_subcription_file(image, request.subsidiary)
         att["acc"]["feedback"], fb_max_indexes = acc_maximize_list_values(att["acc"]["feedback"])
         att["acc"]["reviewed"], rv_max_indexes = acc_maximize_list_values(att["acc"]["reviewed"])
 
-        _att = copy.deepcopy(att)
+        _att = copy.deepcopy(att) # deep copy right here to advoid removing acc for bad images in the next steps
 
         fb_avg_acc = avg_dict(att["acc"]["feedback"])
         rv_avg_acc = avg_dict(att["acc"]["reviewed"])
@@ -924,6 +920,8 @@ def calculate_a_request(report, request):
             continue
         if status != 200:
             continue
+
+        atts.append(att)
         image.feedback_accuracy = att["acc"]["feedback"] # dict {key: [values]}
         image.is_bad_image_quality = att["is_bad_image"] # is_bad_image=avg_acc<threshold (avg_acc=feedback_acc)
         
@@ -1033,7 +1031,7 @@ def calculate_a_request(report, request):
  
     return request_att, report_files, atts
 
-def calculate_subcription_file(subcription_request_file):
+def calculate_subcription_file(subcription_request_file, subsidiary):
     att = {"acc": {"feedback": {},
                     "reviewed": {}},
             "normalized_data": {"feedback": {},
@@ -1050,7 +1048,7 @@ def calculate_subcription_file(subcription_request_file):
     feedback_result = copy.deepcopy(subcription_request_file.feedback_result)
     reviewed_result = copy.deepcopy(subcription_request_file.reviewed_result)
 
-    accuracy_keys_for_this_image = KEYS_BY_FILE_TYPE.get(subcription_request_file.doc_type, [])
+    accuracy_keys_for_this_image = settings.FIELDS_BY_SUB.get(subsidiary, settings.FIELDS_BY_SUB["default"]).get(subcription_request_file.doc_type, [])
 
     for key_name in VALID_KEYS:
         att["acc"]["feedback"][key_name] = []