def apply_24_net(image,iname,nn,bb):
X24 = []
X12 = []
for c,e in enumerate(bb):
x1,y1,x2,y2 = e
sub = image[y1:y2,x1:x2]
sub12 = fr.resize_frame(sub,(12,12))
sub24 = fr.resize_frame(sub,(24,24))
X12.append(fr.frame_to_vect(sub12))
X24.append(fr.frame_to_vect(sub24))
X12 = sp.array(X12)
X24 = sp.array(X24)
pred = nn.predict(X=X24,X12=X12)
return bb[pred==1]
def apply_24_net(image, iname, nn, bb):
X24 = []
X12 = []
for c, e in enumerate(bb):
x1, y1, x2, y2 = e
sub = image[y1:y2, x1:x2]
sub12 = fr.resize_frame(sub, (12, 12))
sub24 = fr.resize_frame(sub, (24, 24))
X12.append(fr.frame_to_vect(sub12))
X24.append(fr.frame_to_vect(sub24))
X12 = sp.array(X12)
X24 = sp.array(X24)
pred = nn.predict(X=X24, X12=X12)
return bb[pred == 1]
def apply_24_48_net(image,iname,nn,bb,debug=0):
X = []
k = int(nn.nn_name[:2])
for c,e in enumerate(bb):
x1,y1,x2,y2 = e
sub = image[y1:y2,x1:x2]
subk = fr.resize_frame(sub,(k,k))
X.append(fr.frame_to_vect(subk))
X = sp.array(X)
pred = nn.predict(X=X)
proba = nn.predict_proba(X=X)
pos_proba = proba[pred==1][:,1]
bb = bb[pred==1]
pos_pred = pred[pred==1]
if nn.nn_name[:2] == '48':
bb = bb[pos_proba > 0.999998]
pred = pred[pos_proba > 0.999998]
pos_proba = pos_proba[pos_proba > 0.999998]
if debug:
j = 0
for e,lb,p in zip(bb,pos_pred,pos_proba):
x1,y1,x2,y2 = e
sub = image[y1:y2,x1:x2]
fr.write_frame('F:\\2\\'+str(j)+ '-'+str(lb) + '-' +str(p) + '-',sub)
j += 1
return (bb,pos_proba)
def apply_calib_net(image, iname, nn, bb):
X_calib = []
for c, e in enumerate(bb):
x1, y1, x2, y2 = e
sub = image[y1:y2, x1:x2]
v = int(nn.nn_name[:2])
subr = fr.resize_frame(sub, (v, v))
X_calib.append(fr.frame_to_vect(subr))
#fr.write_frame('F:\\1\\'+str(c),sub)
X_calib = sp.array(X_calib)
pred_proba = nn.predict_proba(X_calib)
nbb = []
for e, lb in zip(bb, pred_proba):
t = 0.3
x1, y1, x2, y2 = e
sn, xn, yn = mean_pattern(lb, t)
h = abs(y1 - y2)
w = abs(x1 - x2)
#ii,jj,hh,ww = [int(e) for e in calib(x1,y1,h,w,lb)]
#if sp.any(sp.array([ii,jj,hh,ww]) < 0):
# continue
ii, jj, hh, ww = [
max(0, int(e)) for e in calib_apply(x1, y1, h, w, sn, xn, yn)
]
#print(ii,jj,hh,ww)
nbb.append([ii, jj, ii + ww, jj + hh])
for c, e in enumerate(nbb):
x1, y1, x2, y2 = e
sub = image[y1:y2, x1:x2]
#fr.write_frame('F:\\1\\'+str(c) + 'calib',sub)
return sp.array(nbb)
def get_aflw_face_data(k = 12):
dbpath = 'F:\\datasets\\image_data_sets\\faces\\AFLW'
dbpath = join(dbpath,'aflw.sqlite')
rfpath = 'F:\\datasets\\image_data_sets\\faces\\AFLW\\img'
conn = sqlite3.connect(dbpath)
X = []
for file_id,x,y,h,w in conn.execute('SELECT file_id,x,y,h,w FROM Faces NATURAL JOIN FaceRect'):
fpath = join(rfpath,file_id)
frame = fr.get_frame(fpath)
no_neg = sp.all(sp.array([x,y,h,w]) > 0) ## ignore a bad data in sql table
if frame != None and no_neg:
face = fr.get_patch(frame,y,x,(h,w))
face_r,good_example = Datasets.sample_resize(face,k)
if good_example:
print('face:',fpath)
vec = fr.frame_to_vect(face_r)
X.append(vec)
face_flip = fr.flip_frame(face)
face_flip_r = fr.resize_frame(face_flip,(k,k))
vec = fr.frame_to_vect(face_flip_r)
X.append(vec)
#fr.write_frame('F:\\1\\'+'flip'+str(file_id),face_flip)
#fr.write_frame('F:\\1\\'+str(file_id),face)
X = sp.array(X)
y = sp.ones(len(X))
return X,y
def apply_calib_net(image,iname,nn,bb):
X_calib = []
for c,e in enumerate(bb):
x1,y1,x2,y2 = e
sub = image[y1:y2,x1:x2]
v = int(nn.nn_name[:2])
subr = fr.resize_frame(sub,(v,v))
X_calib.append(fr.frame_to_vect(subr))
#fr.write_frame('F:\\1\\'+str(c),sub)
X_calib = sp.array(X_calib)
pred_proba = nn.predict_proba(X_calib)
nbb = []
for e,lb in zip(bb,pred_proba):
t = 0.3
x1,y1,x2,y2 = e
sn,xn,yn = mean_pattern(lb,t)
h = abs(y1-y2)
w = abs(x1-x2)
#ii,jj,hh,ww = [int(e) for e in calib(x1,y1,h,w,lb)]
#if sp.any(sp.array([ii,jj,hh,ww]) < 0):
# continue
ii,jj,hh,ww = [max(0,int(e)) for e in calib_apply(x1,y1,h,w,sn,xn,yn)]
#print(ii,jj,hh,ww)
nbb.append([ii,jj,ii+ww,jj+hh])
for c,e in enumerate(nbb):
x1,y1,x2,y2 = e
sub = image[y1:y2,x1:x2]
#fr.write_frame('F:\\1\\'+str(c) + 'calib',sub)
return sp.array(nbb)
def frames2batch(k=12, batch_size=1024, is_calib=False):
pos = util.get_files(rootdir='F:\\train_data\\pos\\')
neg = util.get_files(rootdir='F:\\train_data\\neg\\')
pos = shuffle(pos)
neg = shuffle(neg)
total = pos + neg
total = shuffle(total)
batch = []
c = 0
bpath = 'F:\\train_data\\batch\\'
for item_path in total:
frame = fr.get_frame(item_path)
frame_r = fr.resize_frame(frame, (k, k))
if frame_r == None:
continue
vec = fr.frame_to_vect(frame_r)
label = 1 if item_path.split('\\')[-1].find('pos') > 0 else 0
print(item_path, label)
batch.append((vec, label))
if len(batch) > 0 and len(batch) % batch_size == 0:
batch = sp.array(batch)
sp.savez(
bpath + str(c) + '_' + str(k) +
('_' if not is_calib else '_calib-') + 'net', batch)
batch = []
c += 1
if len(batch) > 0 and len(batch) % batch_size == 0:
batch = sp.array(batch)
sp.savez(
bpath + str(c) + '_' + str(k) +
('_' if not is_calib else '_calib') + '-net', batch)
batch = []
c += 1
def get_aflw_face_data(k=12):
dbpath = 'F:\\datasets\\image_data_sets\\faces\\AFLW'
dbpath = join(dbpath, 'aflw.sqlite')
rfpath = 'F:\\datasets\\image_data_sets\\faces\\AFLW\\img'
conn = sqlite3.connect(dbpath)
X = []
for file_id, x, y, h, w in conn.execute(
'SELECT file_id,x,y,h,w FROM Faces NATURAL JOIN FaceRect'):
fpath = join(rfpath, file_id)
frame = fr.get_frame(fpath)
no_neg = sp.all(
sp.array([x, y, h, w]) > 0) ## ignore a bad data in sql table
if frame != None and no_neg:
face = fr.get_patch(frame, y, x, (h, w))
face_r, good_example = Datasets.sample_resize(face, k)
if good_example:
print('face:', fpath)
vec = fr.frame_to_vect(face_r)
X.append(vec)
face_flip = fr.flip_frame(face)
face_flip_r = fr.resize_frame(face_flip, (k, k))
vec = fr.frame_to_vect(face_flip_r)
X.append(vec)
#fr.write_frame('F:\\1\\'+'flip'+str(file_id),face_flip)
#fr.write_frame('F:\\1\\'+str(file_id),face)
X = sp.array(X)
y = sp.ones(len(X))
return X, y
def frames2batch(k = 12,batch_size = 1024, is_calib = False):
pos = util.get_files(rootdir = 'F:\\train_data\\pos\\')
neg = util.get_files(rootdir = 'F:\\train_data\\neg\\')
pos = shuffle(pos)
neg = shuffle(neg)
total = pos + neg
total = shuffle(total)
batch = []
c = 0
bpath = 'F:\\train_data\\batch\\'
for item_path in total:
frame = fr.get_frame(item_path)
frame_r = fr.resize_frame(frame,(k,k))
if frame_r == None:
continue
vec = fr.frame_to_vect(frame_r)
label = 1 if item_path.split('\\')[-1].find('pos') > 0 else 0
print(item_path,label)
batch.append((vec,label))
if len(batch) > 0 and len(batch) % batch_size == 0:
batch = sp.array(batch)
sp.savez(bpath + str(c) + '_' + str(k) + ('_' if not is_calib else '_calib-') + 'net',batch)
batch = []
c += 1
if len(batch) > 0 and len(batch) % batch_size == 0:
batch = sp.array(batch)
sp.savez(bpath + str(c) + '_' + str(k) + ('_' if not is_calib else '_calib') + '-net',batch)
batch = []
c += 1
def sample_resize(subframe,k = 12):
'''
return resized frame or reject it if any of side less than k
'''
H,W,dim = subframe.shape
if H >= k and W >= k:
return (fr.resize_frame(subframe,(k,k)),True)
else:
return (subframe,False)
def sample_resize(subframe, k=12):
'''
return resized frame or reject it if any of side less than k
'''
H, W, dim = subframe.shape
if H >= k and W >= k:
return (fr.resize_frame(subframe, (k, k)), True)
else:
return (subframe, False)
def sample_resize(subframe, k_check=12, k_res=12):
'''
return resized frame or reject it if any of side less than k
'''
H, W, dim = subframe.shape
# there are negative coord's in datasets :(
if H <= 0 or W <= 0:
return (subframe, False)
#print(H,W,dim)
if H >= k_check or W >= k_check:
return (fr.resize_frame(subframe, (k_res, k_res)), True)
else:
return (subframe, False)
def sample_resize(subframe,k_check = 12, k_res = 12):
'''
return resized frame or reject it if any of side less than k
'''
H,W,dim = subframe.shape
# there are negative coord's in datasets :(
if H <= 0 or W <= 0:
return (subframe,False)
#print(H,W,dim)
if H >= k_check or W >= k_check:
return (fr.resize_frame(subframe,(k_res,k_res)),True)
else:
return (subframe,False)
def apply_calib_net(image,iname,nn,bb,T=0.3):
X_calib = []
for c,e in enumerate(bb):
x1,y1,x2,y2 = e
sub = image[y1:y2,x1:x2]
v = int(nn.nn_name[:2])
subr = fr.resize_frame(sub,(v,v))
X_calib.append(fr.frame_to_vect(subr))
#fr.write_frame('F:\\1\\'+str(c),sub)
X_calib = sp.array(X_calib)
pred_proba = nn.predict_proba(X_calib)
nbb = []
for e,lb in zip(bb,pred_proba):
t = 1/45
x1,y1,x2,y2 = e
ds, dx, dy = mean_pattern(lb,t)
h = abs(y1-y2)
w = abs(x1-x2)
ii,jj,hh,ww = [int(max(0,e)) for e in util.calib_apply(x1,y1,h,w,ds,dx,dy)]
nbb.append([ii,jj,ii+ww,jj+hh])
return sp.array(nbb)
