def run(self, img, features, bboxes): boxSet = bboxes areas = [det.area(b) for b in boxSet] rank = np.argsort(areas)[::-1] rankedBoxes = [boxSet[i].tolist() for i in rank[0:self.maxTime]] scores = [features[i,self.categoryIndex].tolist() for i in rank[0:maxTime]] return (img, rankedBoxes, scores, range(len(rankedBoxes)))
def run(self, img, features, bboxes):
boxSet = bboxes
areas = [det.area(b) for b in boxSet]
rank = np.argsort(areas)[::-1]
rankedBoxes = [boxSet[i].tolist() for i in rank[0:self.maxTime]]
scores = [
features[i, self.categoryIndex].tolist() for i in rank[0:maxTime]
]
return (img, rankedBoxes, scores, range(len(rankedBoxes)))
def loadGroundTruth(self, imageName):
try:
gt = self.groundTruth[imageName]
except:
gt = []
if self.image != imageName:
self.displayEpisodePerformance()
self.image = imageName
self.boxes = [b[:] for b in gt]
self.centers = [center(b) for b in gt]
self.areas = [det.area(b) for b in gt]
self.control = {'IOU': [0.0 for b in gt], 'DONE': [False for b in gt],
'SKIP': [False for b in gt]}
return gt
def loadFilteredFeaturesAndIndex(img, content, index):
import libDetection as det
import numpy as np
MAX_BOXES = 1000
m, l = loadFeaturesAndIndex(img, content, index)
if len(l) > MAX_BOXES:
limit = max(len(l) - MAX_BOXES, 0)
areas = [det.area(map(float, b[1:])) for b in l]
areasIdx = np.argsort(areas)
candidates = np.asarray([i >= limit for i in areasIdx])
boxes = [l[i] for i in areasIdx if i >= limit]
return (m[candidates].astype(floatType), boxes)
else:
return m, l
def loadFilteredFeaturesAndIndex(img,content,index):
import libDetection as det
import numpy as np
MAX_BOXES = 1000
m,l = loadFeaturesAndIndex(img,content,index)
if len(l) > MAX_BOXES:
limit = max(len(l)-MAX_BOXES,0)
areas = [ det.area(map(float,b[1:])) for b in l]
areasIdx = np.argsort(areas)
candidates = np.asarray([i >= limit for i in areasIdx])
boxes = [l[i] for i in areasIdx if i >= limit]
return (m[candidates].astype(floatType),boxes)
else:
return m,l
def loadGroundTruth(self, imageName):
try:
gt = self.groundTruth[imageName]
except:
gt = []
if self.image != imageName:
self.displayEpisodePerformance()
self.image = imageName
self.boxes = [b[:] for b in gt]
self.centers = [center(b) for b in gt]
self.areas = [det.area(b) for b in gt]
self.control = {
'IOU': [0.0 for b in gt],
'DONE': [False for b in gt],
'SKIP': [False for b in gt]
}
return gt
def __init__(self, box, id): self.id = id self.box = box self.area = det.area(box)
def selectProposalsByArea(proposals,minArea,maxArea):
areas = [det.area(box) for box in proposals]
proposals = [[0]+proposals[i] for i in range(len(areas)) if areas[i] >= minArea and areas[i] < maxArea]
return proposals
import os,sys
import libDetection as det
import masks as mk
if len(sys.argv) < 3:
print 'Use: filterDPMBoxes.py input output'
sys.exit()
boxes = mk.loadBoxIndexFile(sys.argv[1])
totalBoxes = 0
keptBoxes = 0
out = open(sys.argv[2],'w')
for img in boxes.keys():
unique = set([':'.join(map(str,map(int,x))) for x in boxes[img]])
for b in boxes[img]:
boxHash = ':'.join(map(str,map(int,b)))
if boxHash in unique:
unique.remove(boxHash)
else:
continue
totalBoxes += 1
a = det.area(b)
w,h = b[2]-b[0],b[3]-b[1]
if a > 400 and w > 0 and h > 0:
out.write(img + ' ' + ' '.join(map(str,map(int,b))) + '\n' )
keptBoxes += 1
print 'Total boxes:',totalBoxes,'Filtered:',totalBoxes-keptBoxes,'Final Boxes:',keptBoxes
out.close()
def selectProposalsByArea(proposals, minArea, maxArea):
areas = [det.area(box) for box in proposals]
proposals = [[0] + proposals[i] for i in range(len(areas))
if areas[i] >= minArea and areas[i] < maxArea]
return proposals
def __init__(self, box, id): self.id = id self.box = box self.area = det.area(box)
import os,sys
import utils as cu
import libDetection as det
import numpy as np
if __name__ == "__main__":
params = cu.loadParams('boxesFile minSize outputFile')
boxes = [x.split() for x in open(params['boxesFile'])]
minA = float(params['minSize'])**2
images = {}
found = set()
for box in boxes:
a = det.area( map(int,box[1:]) )
if a >= minA:
try:
images[ box[0] ].append(box[1:])
except:
images[ box[0] ] = [box[1:]]
found.add(box[0])
# Add records for images that do not have enough area to comply with the filter
missing = found.symmetric_difference( images.keys() )
missing = [b for b in boxes if b[0] in missing]
allAreas = {}
for m in missing:
img = m[0]
try:
allAreas[img]['box'].append( m[1:] )
allAreas[img]['area'].append( det.area( map(int,m[1:]) ) )
except:
allAreas[img] = { 'box':[m[1:]], 'area':[det.area( map(int,m[1:]) )] }
for img in allAreas.keys():