first pass/thoughts on faces per file in DB, mostyl commented out. Also put a poor hack in to allow AI jobs to re-run without error, its limited, need to really fix this properly

pa_job_manager.py

@@ -103,6 +103,9 @@ class File(Base):
     size_mb = Column(Integer, unique=False, nullable=False)
     hash = Column(Integer, unique=True, nullable=True)
     thumbnail = Column(String, unique=False, nullable=True)
+#   DDP: need bytea? in db (see other DDP comment)
+#    faces = 
+    faces_created_on = Column(Float)
 
     def __repr__(self):
         return "<eid: {}, size_mb={}, hash={}>".format(self.eid, self.size_mb, self.hash )
@@ -478,6 +481,10 @@ def JobImportDir(job):
     return
 
 def JobProcessAI(job):
+    print ("DDP: HACK - to allow re-running jobs for now, del FPL");
+    session.query(File_Person_Link).delete()
+    #### (delete the above 2 lines)
+
     path=[jex.value for jex in job.extra if jex.name == "path"][0]
     path = SymlinkName(path, '/')
     print('REMOVE AFTER TESTING ON WINDOWS... path=',path)
@@ -512,11 +519,26 @@ def ProcessAI(job, e):
         generateKnownEncodings(person)
     
     file = e.in_dir[0].path_prefix + '/' + e.name
-    im_orig = Image.open(file)
-    im = ImageOps.exif_transpose(im_orig)
-    unknown_encodings = generateUnknownEncodings(im)
+    stat = os.stat(file)
+    # only find faces if we have not already OR file is newer than when we found faces before
+    if not e.file_details[0].faces_created_on or stat.st_ctime > e.file_details[0].faces_created_on:
+        session.add(e)
+        im = Image.open(file)
+        try:
+            im = ImageOps.exif_transpose(im)
+        except:
+            print("DEBUG: looks like image does not have exif")
 
-    for unknown_encoding in unknown_encodings:
+        faces = generateUnknownEncodings(im)
+#       DDP: uncomment the below to optimise, but I need to store the faces into the DB, not sure how right now
+##### is this really 0? or will there be many with the many faces?
+#       if its many, should we do a faces_file_link???
+#        e.file_details[0].faces = faces[0].tobytes()
+#        e.file_details[0].faces_created_on=time.time()
+#   else:
+#       faces=numpy.frombuffer(e.file_details[0].faces,dtype=numpy.float64)      
+
+    for unknown_encoding in faces:
         for person in people:
             lookForPersonInImage(job, person, unknown_encoding, e)
     AddLogForJob(job, f"Finished processing {e.name}", e.name )
@@ -524,9 +546,11 @@ def ProcessAI(job, e):
 
 def lookForPersonInImage(job, person, unknown_encoding, e):
     for refimg in person.refimg:
+        ###
+        # need a date_stamp in refimg_file_link, but we currently have a person_file_link
+        #   should consider whether we break this into just a scan ( id, refimg, file, date, threshold, etc.)
+        ###
         deserialized_bytes = numpy.frombuffer(refimg.encodings, dtype=numpy.float64)
-        #deserialized_x = numpy.reshape(deserialized_bytes, newshape=(2,2))
-
         results = compareAI(deserialized_bytes, unknown_encoding)
         if results[0]:
             print(f'Found a match between: {person.tag} and {e.name}')

tables.sql

@@ -8,7 +8,7 @@ create table ENTRY( ID integer, NAME varchar(128), TYPE_ID integer,
     constraint PK_ENTRY_ID primary key(ID),
     constraint FK_FILE_TYPE_TYPE_ID foreign key (TYPE_ID) references FILE_TYPE(ID) );
 
-create table FILE ( EID integer, SIZE_MB integer, HASH varchar(34), THUMBNAIL varchar, 
+create table FILE ( EID integer, SIZE_MB integer, HASH varchar(34), THUMBNAIL varchar, FACES_CREATED_ON float,
     constraint PK_FILE_ID primary key(EID), 
     constraint FK_FILE_ENTRY_ID foreign key (EID) references ENTRY(ID) );