|
|
|
@ -6,62 +6,6 @@ from optparse import OptionParser |
|
|
|
|
import re |
|
|
|
|
import numpy as np |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def extract(f): |
|
|
|
|
person = re.compile("^lbl=\'person\'\s+str=(\d+)\s+end=(\d+)\s+hide=0$") |
|
|
|
|
newobj = re.compile("^lbl=\'(\w+)\'\s+str=(\d+)\s+end=(\d+)\s+hide=0$") |
|
|
|
|
pos = re.compile("^pos\s=(\[[((\d+\.+\d*)|\s+|\;)]*\])$") |
|
|
|
|
nonarray = re.compile("\;\s+(?!\])|\s+(?!\])") |
|
|
|
|
lastSemicolon = re.compile("\;\s+(?=\])") |
|
|
|
|
|
|
|
|
|
goNext = 0 |
|
|
|
|
start = 0 |
|
|
|
|
end = 0 |
|
|
|
|
|
|
|
|
|
modelW = 32 |
|
|
|
|
modelH = 64 |
|
|
|
|
|
|
|
|
|
for l in f: |
|
|
|
|
qq = newobj.match(l) |
|
|
|
|
if qq is not None: |
|
|
|
|
if qq.group(1) == "person": |
|
|
|
|
goNext = 1 |
|
|
|
|
else: |
|
|
|
|
goNext = 0 |
|
|
|
|
print qq.group(0), qq.group(1) |
|
|
|
|
m = person.match(l) |
|
|
|
|
if m is not None: |
|
|
|
|
start = m.group(1) |
|
|
|
|
end = m.group(2) |
|
|
|
|
|
|
|
|
|
print m.group(0), start, end |
|
|
|
|
else: |
|
|
|
|
m = pos.match(l) |
|
|
|
|
if m is not None: |
|
|
|
|
if not goNext: |
|
|
|
|
continue |
|
|
|
|
strarr = re.sub(r"\s", ", ", re.sub(r"\;\s+(?=\])", "]", re.sub(r"\;\s+(?!\])", "],[", re.sub(r"(\[)(\d)", "\\1[\\2", m.group(1))))) |
|
|
|
|
list = eval(strarr) |
|
|
|
|
for idx, box in enumerate(list): |
|
|
|
|
if (box[2] >= 32) or (box[3] >= 64): |
|
|
|
|
x = box[0] |
|
|
|
|
y = box[1] |
|
|
|
|
w = box[2] |
|
|
|
|
h = box[3] |
|
|
|
|
|
|
|
|
|
ratio = w / h |
|
|
|
|
neww = h / 2.0 |
|
|
|
|
offset = (w - neww) / 2.0 |
|
|
|
|
print "HERE is big!! ", box, ratio, offset |
|
|
|
|
if (x + offset) > 0: |
|
|
|
|
id = int(start) + idx |
|
|
|
|
file = "/home/kellan/datasets/caltech/set00/V004.seq/I0%04d.jpg" % id # I00000.jpg |
|
|
|
|
print file |
|
|
|
|
img = cv2.imread(file) |
|
|
|
|
cv2.rectangle(img, (int(x), int(y)), (int(x + w), int(y + h)), (0,255,0), 2) |
|
|
|
|
cv2.imshow("sample", img) |
|
|
|
|
cv2.waitKey(50) |
|
|
|
|
|
|
|
|
|
def showPeople(f, path, opath): |
|
|
|
|
newobj = re.compile("^lbl=\'(\w+)\'\s+str=(\d+)\s+end=(\d+)\s+hide=0$") |
|
|
|
|
pos = re.compile("^pos\s=(\[[((\d+\.+\d*)|\s+|\;)]*\])$") |
|
|
|
@ -152,20 +96,9 @@ if __name__ == "__main__": |
|
|
|
|
opath = os.path.join(options.output, datetime.now().strftime("raw_ge48-" + "-%Y-%m-%d-%H-%M-%S")) |
|
|
|
|
os.mkdir(opath) |
|
|
|
|
|
|
|
|
|
# mat = cv2.imread("/home/kellan/datasets/INRIArescaled/training_set/pos/octave_-1/sample_0.png"); |
|
|
|
|
# cv2.rectangle(mat, (10, 10), (42, 74), (8, 107, 255), 1) |
|
|
|
|
|
|
|
|
|
# cv2.imshow("person", mat) |
|
|
|
|
# cv2.waitKey(0) |
|
|
|
|
# if c == 27: |
|
|
|
|
# exit(0) |
|
|
|
|
|
|
|
|
|
gl = glob.iglob( os.path.join(options.input, "set[0-1][0-9]/V0[0-9][0-9].txt")) |
|
|
|
|
for each in gl: |
|
|
|
|
path, ext = os.path.splitext(each) |
|
|
|
|
path = path + ".seq" |
|
|
|
|
print path |
|
|
|
|
showPeople(open(each), path, opath) |
|
|
|
|
|
|
|
|
|
# f = open("/home/kellan/datasets/caltech/set00/V004.txt") |
|
|
|
|
# extract(f) |
|
|
|
|
showPeople(open(each), path, opath) |