def inference(feature_list, task, img):
x = -1
name = ""
dist = 0
for x in [0, 80, 160]:
for y in [0, 60, 120]:
print(x, y)
gc.collect()
#img2 = img.copy((x,y,160,120))
img2 = img.copy((0, 0, 320, 240))
fmap = kpu.forward(task, img2)
#p = np.array(fmap[:])
kpu.fmap_free(fmap)
'''
# get nearest target
name,dist,v = get_nearest(feature_list, p)
print("name:" + name + " dist:" + str(dist) + " v:" + str(v) + " mem:" + str(gc.mem_free()))
if dist < 200:
break
if dist < 200:
break
'''
#img2 = img.copy((x,y,160,120))
fmap = kpu.forward(task, img)
p = np.array(fmap[:])
kpu.fmap_free(fmap)
# get nearest target
name, dist, v = get_nearest(feature_list, p)
print("name:" + name + " dist:" + str(dist) + " v:" + str(v) + " mem:" +
str(gc.mem_free()))
print(x)
return name, dist, x
def mnist_run(img, dx, dy, dis, x00=0, y00=80, nnn=2):
if nnn == 4:
x00 = x00
dy = dy
img0 = img.copy((x00 + dis * nnn, y00 + nnn * 0, dx, dy))
#img0.mean(2, threshold=True, offset=1, invert=True) #A
img0.median(2, percentile=0.3, threshold=True, offset=-3, invert=True)
#img0.midpoint(2, bias=0.3, threshold=True, offset=0, invert=True)
#img0.mode(2, threshold=True, offset=0, invert=True) #B
#img0.binary([(110,255)], invert = True)
for dx0 in range(dx):
for dy0 in range(dy):
a0 = img0.get_pixel(dx0, dy0)
img.set_pixel(x00 + dis * nnn + dx0, y00 + nnn * 0 + dy0, a0)
#img1 = img0.copy((1,1, dx-1, dy-1))
img1 = img0
img1 = img1.resize(28, 28)
img1 = img1.to_grayscale(1)
img1.pix_to_ai()
fmap = kpu.forward(task, img1)
plist = fmap[:]
pmax = max(plist)
max_index = plist.index(pmax)
kpu.fmap_free(fmap)
return max_index, pmax
def wizard(task):
delay(1, '0')
play_sound("/sd/oshaku/start_learn.wav")
free()
delay(1500, '0')
for ang in ["0","45","90","135","non"]:
play_sound("/sd/oshaku/" + str(ang) + ".wav")
delay(1000, ang)
for i in ["3","2","1"]:
play_sound("/sd/oshaku/" + str(i) + ".wav")
delay(700, ang)
play_sound("/sd/oshaku/kacha.wav")
img = sensor.snapshot()
feature = get_feature(task, img)
free()
feature_list.append([ang,feature])
save(feature_file, feature_list)
drawHeader(img, "Point the camera")
drawAngle(img, ang)
drawFooter(img, "Saved " + ang + " degrees")
lcd.display(img)
free()
play_sound("/sd/oshaku/set.wav")
kpu.fmap_free(feature)
feature_100 = []
save(feature_file, feature_list)
br.play_sound("/sd/reset.wav")
else:
br.play_sound("/sd/camera.wav")
feature = get_feature(task)
feature_list.append([name,feature[:]])
if name=='0':
feature_0.append([name,feature[:]])
if name=='100':
feature_100.append([name,feature[:]])
save(feature_file, feature_list)
br.play_sound("/sd/set.wav")
gc.collect()
# print(gc.mem_free())
kpu.fmap_free(feature)
print("[QR]: " + name)
continue
# inference
fmap = kpu.forward(task, img)
plist=fmap[:]
clock.tick()
if len(feature_0)>0 and len(feature_100)>0:
p = plist
f0 = feature_0[0]
f100 = feature_100[0]
dist = 100.0 * get_dist(f0[1],f100[1],p)
dist_str = "%.1f"%(dist)
print("[DISTANCE]: " + dist_str)
img.draw_string(2,47, dist_str,scale=3)
try:
while (True):
img = sensor.snapshot()
fmap = kpu.forward(task, img)
plist = fmap[:]
dist = 0
for i in range(768):
dist = dist + (plist[i] - firstdata[i]) * (plist[i] - firstdata[i])
if dist < 200:
img.draw_rectangle(1, 46, 222, 132, color=31, thickness=3)
img.draw_string(2, 47, "%.2f " % (dist))
lcd.display(img)
dist_int = int(dist)
print(dist_int)
img_size1 = (dist_int & 0xFF0000) >> 16
img_size2 = (dist_int & 0x00FF00) >> 8
img_size3 = (dist_int & 0x0000FF) >> 0
data_packet = bytearray([
0xFF, 0xD8, 0xEA, 0x01, img_size1, img_size2, img_size3, 0x00,
0x00, 0x00
])
uart_Port.write(data_packet)
if but_b.value() == 0:
firstmap = kpu.forward(task, img)
firstdata = firstmap[:]
kpu.fmap_free(fmap)
except KeyboardInterrupt:
kpu.deinit(task)
sys.exit()
color=(0, 144, 255),
thickness=10)
img.draw_string(50,
55,
"Class:%d" % (class_num, ),
color=(255, 255, 255),
scale=1)
lcd.display(img)
img.pix_to_ai()
#print(img)
feature = get_feature(task, img)
del img
l, qvec = quantize_vector(feature[:])
feature_list.append((l, qvec, class_num))
gc.collect()
kpu.fmap_free(feature)
lcd.clear()
for l1, qvec1, class_num in feature_list:
similarities = ""
for l2, qvec2, _ in feature_list:
dist = get_cos_distance(l1, qvec1, l2, qvec2)
similarities += "%0.2f " % (dist, )
print(class_num, similarities)
sp_device = None
if should_connect_spike_prime:
show_message("Connecting to LPF2 Hub...",
x=100,
bg_color=lcd.BLACK)