security.py

class Frame:
    # Update this class for every components used in the calculation
    __slots__ = ('act_labs', 'act_confs', 'dobj', 'face_names', 'level')
    
    # Recognized gestures counted as negative
    NEGLABS = { "barang2_DR", "barang2_UR", "barang2_DL", "barang2_UL",
                "barang1l_DR", "barang1l_UR", "barang1l_DL", "barang1l_UL",
                "barang1r_DR", "barang1r_UR", "barang1r_DL", "barang1r_UL"}
    # NEGLABS = { "barang_NE", "barang_NW", "barang_SE", "barang_SW"}
    NEGOBJ = {"something"}

    def __init__(self, act_labs, act_confs, dobj, face_names):
        self.act_labs = act_labs
        self.act_confs = act_confs
        self.dobj = dobj
        self.face_names = face_names
        self.level = -1     # Uncalculated value
        
    def calc(self):
        # The calculation depth is basically endless.
        # Some ideas to consider:
        # - Pairing obj location with person/pose
        # - Check face detection with detected person/pose
        # - Positive behaviors
        
        all = 0    # Count of total componenets
        neg = 0    # Count of negative componenets
        
        for (lab, conf) in zip(self.act_labs, self.act_confs):
            if lab in self.NEGLABS:
                neg += 1
            elif conf < 0.2:
                neg += 1
            all += 1
        
        # Disable all+=1 if you don't want other
        # positive recog to skew the gesture result
        
        # for obj in self.dobj:
            # # An array: obj = [label, confidence, bounds]
            # if obj[0] in self.NEGOBJ:
                # neg += 1
            # all += 1
            
        # Disable if unknown face is permitted
        # for face in self.face_names:
            # if face == "Unknown":
                # neg += 1
            # all += 1
        
        if all == 0:
            all = 1
            
        if neg > all:
            neg = all
        
        # if all == 0:
            # # Force 0.5
            # all = 2
            # neg = 1
        
        # Get security conclusion of a single frame
        self.level = (all-neg)/all
        
                    
        # Count face unknown percentage
        # for face in self.face_names:
            # if face == "Unknown":
                # fneg += 1
            # fall += 1
            
        # if fall == 0:
            # fall = 1
            
        # self.flevel = (fall-fneg)/fall
        
    # def __init__(self, pairs):
        # self.pairs = []
        # self.uidx_list = set()
        # self.body_parts = {}
        # for pair in pairs:
            # self.add_pair(pair)
        # self.score = 0.0