Beispiel #1
0
def clustercost(data,objectDict,baseline = 0.001):
    #    data is a tuple of two dictionaries: core cluster data first, then larger and more permissive clusters\
    # this func needs to return a unified list of possible clusters using both dictionaries in the style of the chain finder function
    # quick and dirty:
    
    smallClusters = []
    bigClusters = []

    
    
    
    
    #cores
    for i in data[0].values():
        corePO = map(lambda x: objectDict.get(x),i)
        coreHull = cluster_util.convex_hull(corePO)
        try:
            density = len(i)/cluster_util.area(coreHull)
        except:
            density = 0
        if density>=1: print "groups are too dense, things may be weird."
        cost = len(i)*(1-density) + baseline
        
        corecluster=cluster_util.GroupBundle(i,cost)
        smallClusters.append(corecluster)
        
    #fringes
    for i in data[1].values():
        fringePO = map(lambda x: objectDict.get(x),i)
        fringeHull = cluster_util.convex_hull(fringePO)
        try:
            density = len(i)/cluster_util.area(fringeHull)
        except:
            density = 0
        if density>=1: print "groups are too dense, things may be weird."
        cost = len(i)*(1-density) + baseline
    
        bigcluster = cluster_util.GroupBundle(i,cost)
        bigClusters.append(bigcluster)
        
    return (smallClusters,bigClusters)
Beispiel #2
0
def clustercost(data, objectDict, baseline=0.001):
    #    data is a tuple of two dictionaries: core cluster data first, then larger and more permissive clusters\
    # this func needs to return a unified list of possible clusters using both dictionaries in the style of the chain finder function
    # quick and dirty:

    smallClusters = []
    bigClusters = []

    #cores
    for i in data[0].values():
        corePO = map(lambda x: objectDict.get(x), i)
        coreHull = cluster_util.convex_hull(corePO)
        try:
            density = len(i) / cluster_util.area(coreHull)
        except:
            density = 0
        if density >= 1: print "groups are too dense, things may be weird."
        cost = len(i) * (1 - density) + baseline

        corecluster = cluster_util.GroupBundle(i, cost)
        smallClusters.append(corecluster)

    #fringes
    for i in data[1].values():
        fringePO = map(lambda x: objectDict.get(x), i)
        fringeHull = cluster_util.convex_hull(fringePO)
        try:
            density = len(i) / cluster_util.area(fringeHull)
        except:
            density = 0
        if density >= 1: print "groups are too dense, things may be weird."
        cost = len(i) * (1 - density) + baseline

        bigcluster = cluster_util.GroupBundle(i, cost)
        bigClusters.append(bigcluster)

    return (smallClusters, bigClusters)
Beispiel #3
0
def clusterCostWorker(data,objectDict,maxDiscountRatio,strong=False):
    #maxDiscountRatio is the maximum amount of the groups cost that can be deducted
    #for membership in a group. Higher values will make groups more favorable compared
    # to lines.
    output = []
    for i in data:
        clusterMembers = map(lambda x: objectDict.get(x),i)
        hullAroundMembers = cluster_util.convex_hull(clusterMembers)
        areaOfMembers = sum(cluster_util.bb_area(objectDict.get(p).representation.rect) for p in i)
        try:
            density = areaOfMembers/cluster_util.area(hullAroundMembers)
        except:
            density = 0
        if density>=1: density = 1
        cost = len(i)*(1-density**2)*maxDiscountRatio + len(i)*(1-maxDiscountRatio)
        
#        corecluster=cluster_util.GroupBundle(i,cost,hullAroundMembers)
 
        corecluster=landmark.GroupRectangleRepresentation([objectDict.get(j) for j in i],None,cost)
        corecluster.density = density
        corecluster.isStrong = strong
        output.append(corecluster)
    return output