def plot_timeVfreq(date_operator, collection):
if collection == "drop":
field = "pickup_time"
elif collection == "pickup":
field = "drop_time"
else:
raise Exception(
"Error: the collection arguments to timeVfreq are 'drop', or 'pickup'. Not",
collection)
graph_size = {"hour": (0, 23), "dayOfYear": (1, 366), "dayOfWeek": (1, 8)}
if date_operator not in graph_size.keys():
raise Exception(
"Error: the date_operator argument to timeVfreq must be 'hour', 'dayOfWeek', or 'dayOfYear'. Not",
date_operator)
tvf = get_timeVfreq(date_operator, collection, field)
min_freq = ndarray.min(tvf[1])
max_freq = ndarray.max(tvf[1])
fig = pyplot.figure(1, figsize=(7, 6))
myplot = fig.add_subplot(111)
myplot.set_xlim(graph_size[date_operator])
norm = colors.Normalize(min_freq, max_freq)
for i in range(len(tvf[0])):
color = colors.rgb2hex(cm.cool(norm(tvf[1][i])))
pyplot.bar(tvf[0][i], tvf[1][i], color=color, ec=color, align='edge')
pyplot.xlabel(date_operator)
pyplot.ylabel("# of rides")
pyplot.savefig(date_operator + "_" + collection, dpi=180)
def plot_timeVfreq(date_operator, collection):
if collection == "drop":
field = "pickup_time"
elif collection == "pickup":
field = "drop_time"
else:
raise Exception("Error: the collection arguments to timeVfreq are 'drop', or 'pickup'. Not", collection)
graph_size = {"hour":(0, 23), "dayOfYear":(1,366), "dayOfWeek":(1,8)}
if date_operator not in graph_size.keys():
raise Exception("Error: the date_operator argument to timeVfreq must be 'hour', 'dayOfWeek', or 'dayOfYear'. Not", date_operator)
tvf = get_timeVfreq(date_operator, collection, field)
min_freq = ndarray.min(tvf[1])
max_freq = ndarray.max(tvf[1])
fig = pyplot.figure(1, figsize=(7, 6))
myplot = fig.add_subplot(111)
myplot.set_xlim(graph_size[date_operator])
norm = colors.Normalize(min_freq, max_freq)
for i in range(len(tvf[0])):
color = colors.rgb2hex(cm.cool(norm(tvf[1][i])))
pyplot.bar(tvf[0][i], tvf[1][i], color=color, ec=color, align='edge')
pyplot.xlabel(date_operator)
pyplot.ylabel("# of rides")
pyplot.savefig(date_operator + "_" + collection, dpi=180)
def max(self, *args, **kwargs):
"""
Returns the maximum Date.
For a description of the input parameters, please refer to numpy.max.
"""
obj = ndarray.max(self, *args, **kwargs)
if not obj.shape:
return Date(self.freq, obj)
return obj
def max(self, *args, **kwargs):
"""
Returns the maximum Date.
For a description of the input parameters, please refer to numpy.max.
"""
obj = ndarray.max(self, *args, **kwargs)
if not obj.shape:
return Date(self.freq, obj)
return obj
def BackgroundImage(request):
chad = User.objects.get(id=1)
image = ThreeDimensional.objects.filter(user=chad)[0]
image_handle = nibabel.load(image.brain_image.file.name)
image_data = image_handle.get_data()
image_list_data = image_data.tolist()
#for some reason ndarray.max returns an ndarray with 1 member
#which is a numpy.flaot32. this converts all that to a regular python float
max = ndarray.max(image_data).tolist()
min = ndarray.min(image_data).tolist()
json_object = {
'data' : image_list_data,
'max' : max,
'min' : min,
}
json_data = json.dumps(json_object)
return HttpResponse(json_data, mimetype='application/json')
def make_validation_plots(opt, tlm, db):
"""
Make validation output plots.
:param outdir: output directory
:param tlm: telemetry
:param db: database handle
:returns: list of plot info including plot file names
"""
outdir = opt.outdir
start = tlm['date'][0]
stop = tlm['date'][-1]
states = get_states(start, stop, db)
# Create array of times at which to calculate PSMC temperatures, then do it
logger.info('Calculating PSMC thermal model for validation')
model = calc_model(opt.model_spec, states, start, stop)
# Interpolate states onto the tlm.date grid
# state_vals = cmd_states.interpolate_states(states, model.times)
pred = {'1pdeaat': model.comp['1pdeaat'].mvals,
'pitch': model.comp['pitch'].mvals,
'tscpos': model.comp['sim_z'].mvals
}
idxs = Ska.Numpy.interpolate(np.arange(len(tlm)), tlm['date'], model.times,
method='nearest')
tlm = tlm[idxs]
labels = {'1pdeaat': 'Degrees (C)',
'pitch': 'Pitch (degrees)',
'tscpos': 'SIM-Z (steps/1000)',
}
scales = {'tscpos': 1000.}
fmts = {'1pdeaat': '%.2f',
'pitch': '%.3f',
'tscpos': '%d'}
good_mask = np.ones(len(tlm),dtype='bool')
for interval in model.bad_times:
bad = ((tlm['date'] >= DateTime(interval[0]).secs)
& (tlm['date'] < DateTime(interval[1]).secs))
good_mask[bad] = False
plots = []
logger.info('Making PSMC model validation plots and quantile table')
quantiles = (1, 5, 16, 50, 84, 95, 99)
# store lines of quantile table in a string and write out later
quant_table = ''
quant_head = ",".join(['MSID'] + ["quant%d" % x for x in quantiles])
quant_table += quant_head + "\n"
for fig_id, msid in enumerate(sorted(pred)):
plot = dict(msid=msid.upper())
fig = plt.figure(10 + fig_id, figsize=(7, 3.5))
fig.clf()
scale = scales.get(msid, 1.0)
ticklocs, fig, ax = plot_cxctime(model.times, tlm[msid] / scale,
fig=fig, fmt='-r')
ticklocs, fig, ax = plot_cxctime(model.times, pred[msid] / scale,
fig=fig, fmt='-b')
if np.any(~good_mask) :
ticklocs, fig, ax = plot_cxctime(model.times[~good_mask], tlm[msid][~good_mask] / scale,
fig=fig, fmt='.c')
ax.set_title(msid.upper() + ' validation')
ax.set_ylabel(labels[msid])
ax.grid()
filename = msid + '_valid.png'
outfile = os.path.join(outdir, filename)
logger.info('Writing plot file %s' % outfile)
fig.savefig(outfile)
plot['lines'] = filename
# Make quantiles
if msid == '1pdeaat':
ok = (( tlm[msid] > 30.0 ) & good_mask )
ok2 =(( tlm[msid] > 40.0 ) & good_mask )
else:
ok = np.ones(len(tlm[msid]), dtype=bool)
diff = np.sort(tlm[msid][ok] - pred[msid][ok])
quant_line = "%s" % msid
for quant in quantiles:
quant_val = diff[(len(diff) * quant) // 100]
plot['quant%02d' % quant] = fmts[msid] % quant_val
quant_line += (',' + fmts[msid] % quant_val)
quant_table += quant_line + "\n"
for histscale in ('log', 'lin'):
fig = plt.figure(20 + fig_id, figsize=(4, 3))
fig.clf()
ax = fig.gca()
ax.hist(diff / scale, bins=50, log=(histscale == 'log'))
if msid == '1pdeaat':
diff2=np.sort(tlm[msid][ok2] - pred[msid][ok2])
ax.hist(diff2 / scale, bins=50, log=(histscale == 'log'),
color = 'red')
ax.set_title(msid.upper() + ' residuals: data - model')
ax.set_xlabel(labels[msid])
fig.subplots_adjust(bottom=0.18)
filename = '%s_valid_hist_%s.png' % (msid, histscale)
outfile = os.path.join(outdir, filename)
logger.info('Writing plot file %s' % outfile)
fig.savefig(outfile)
plot['hist' + histscale] = filename
plots.append(plot)
filename = os.path.join(outdir, 'validation_quant.csv')
logger.info('Writing quantile table %s' % filename)
f = open(filename, 'w')
f.write(quant_table)
f.close()
# If run_start is specified this is likely for regression testing
# or other debugging. In this case write out the full predicted and
# telemetered dataset as a pickle.
if opt.run_start:
filename = os.path.join(outdir, 'validation_data.pkl')
logger.info('Writing validation data %s' % filename)
f = open(filename, 'w')
pickle.dump({'pred': pred, 'tlm': tlm}, f, protocol=-1)
f.close()
# adding stuff for resid plots--rje 6/24/14
fig = plt.figure(36)
fig.clf()
# this is the python equivalent of the IDL where() function
# note parens are required for the & cases.
msid='1pdeaat'
hot_hrcs = ((tlm['tscpos'] < -85000.0 ) & ( pred[msid] > 40.0 ) & good_mask )
hot_hrci = ( ( -85000.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 0.0 ) & ( pred[msid] > 40.0 ) & good_mask )
hot_aciss = ( ( 0.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 80000.0 ) & ( pred[msid] > 40.0 ) & good_mask )
hot_acisi = ((tlm['tscpos'] > 80000.0 ) & ( pred[msid] > 40.0 ) & good_mask )
warm_hrcs = ((tlm['tscpos'] < -85000.0 ) & ( pred[msid] > 30.0 ) & ( pred[msid] < 40.0 ) & good_mask )
warm_hrci = ( ( -85000.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 0.0 )& ( pred[msid] > 30.0 ) & ( pred[msid] < 40.0 ) & good_mask )
warm_aciss = ( ( 0.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 80000.0 )& ( pred[msid] > 30.0 ) & ( pred[msid] < 40.0 ) & good_mask )
warm_acisi = ((tlm['tscpos'] > 80000.0 ) & ( pred[msid] > 30.0 ) & ( pred[msid] < 40.0 ) & good_mask )
cold_hrcs = ( (tlm['tscpos'] < -85000.0 ) & ( pred[msid] < 30.0 ) & good_mask )
cold_hrci = ( ( -85000.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 0.0 ) & ( pred[msid] < 30.0 ) & good_mask )
cold_aciss = ( ( 0.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 80000.0 ) & ( pred[msid] < 30.0 ) & good_mask )
cold_acisi = ( (tlm['tscpos'] > 80000.0 ) & ( pred[msid] < 30.0 ) & good_mask )
plt.plot(tlm['pitch'][hot_hrci], tlm[msid][hot_hrci] - pred[msid][hot_hrci], "ob", markersize=5)
plt.plot(tlm['pitch'][hot_hrcs], tlm[msid][hot_hrcs] - pred[msid][hot_hrcs], "ok", markersize=5)
plt.plot(tlm['pitch'][hot_aciss], tlm[msid][hot_aciss] - pred[msid][hot_aciss], "or", markersize=5)
plt.plot(tlm['pitch'][hot_acisi], tlm[msid][hot_acisi] - pred[msid][hot_acisi], "og", markersize=5)
plt.plot(tlm['pitch'][warm_hrci], tlm[msid][warm_hrci] - pred[msid][warm_hrci], "sb", markersize=3)
plt.plot(tlm['pitch'][warm_hrcs], tlm[msid][warm_hrcs] - pred[msid][warm_hrcs], "sk", markersize=3)
plt.plot(tlm['pitch'][warm_aciss], tlm[msid][warm_aciss] - pred[msid][warm_aciss], "sr", markersize=3)
plt.plot(tlm['pitch'][warm_acisi], tlm[msid][warm_acisi] - pred[msid][warm_acisi], "sg", markersize=3)
plt.plot(tlm['pitch'][cold_hrci], tlm[msid][cold_hrci] - pred[msid][cold_hrci], ".b", markersize=2)
plt.plot(tlm['pitch'][cold_hrcs], tlm[msid][cold_hrcs] - pred[msid][cold_hrcs], ".k", markersize=2)
plt.plot(tlm['pitch'][cold_aciss], tlm[msid][cold_aciss] - pred[msid][cold_aciss], ".r", markersize=2)
plt.plot(tlm['pitch'][cold_acisi], tlm[msid][cold_acisi] - pred[msid][cold_acisi], ".g", markersize=2)
# plt.plot(tlm['pitch'][htr_on], tlm[msid][htr_on] - pred[msid][htr_on], "*m", markersize=10)
plt.ylabel('1PDEAAT Data - Model')
plt.xlabel('pitch angle')
plt.title('b,k,r,g=hrci,hrcs,aciss,acisi, mod.temp: 0<.<30<s<40<o')
plt.grid()
outfile=os.path.join(outdir,'1pdeaat_resid_pitch.png')
fig.savefig(outfile)
# adding stuff for resid plots--rje 6/24/14
fig = plt.figure(35)
fig.clf()
# this is the python equivalent of the IDL where() function
# note parens are required for the & cases.
fwd_hrcs = ((tlm['tscpos'] < -85000.0 ) & ( tlm['pitch'] < 65.0 ) & good_mask )
fwd_hrci = ( ( -85000.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 0.0 ) & ( tlm['pitch'] < 65.0 ) & good_mask )
fwd_aciss = ( ( 0.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 80000.0 ) & ( tlm['pitch'] < 65.0 ) & good_mask )
fwd_acisi = ((tlm['tscpos'] > 80000.0 ) & ( tlm['pitch'] < 65.0 ) & good_mask )
m80_hrcs = ((tlm['tscpos'] < -85000.0 ) & ( tlm['pitch'] > 65.0 ) & ( tlm['pitch'] < 80.0 ) & good_mask )
m80_hrci = ( ( -85000.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 0.0 )& ( tlm['pitch'] > 65.0 ) & ( tlm['pitch'] < 80.0 ) & good_mask )
m80_aciss = ( ( 0.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 80000.0 )& ( tlm['pitch'] > 65.0 ) & ( tlm['pitch'] < 80.0 ) & good_mask )
m80_acisi = ((tlm['tscpos'] > 80000.0 ) & ( tlm['pitch'] > 65.0 ) & ( tlm['pitch'] < 80.0 ) & good_mask )
mid_hrcs = ((tlm['tscpos'] < -85000.0 ) & ( tlm['pitch'] > 80.0 ) & ( tlm['pitch'] < 90.0 ) & good_mask )
mid_hrci = ( ( -85000.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 0.0 )& ( tlm['pitch'] > 80.0 ) & ( tlm['pitch'] < 90.0 ) & good_mask )
mid_aciss = ( ( 0.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 80000.0 )& ( tlm['pitch'] > 80.0 ) & ( tlm['pitch'] < 90.0 ) & good_mask )
mid_acisi = ((tlm['tscpos'] > 80000.0 ) & ( tlm['pitch'] > 80.0 ) & ( tlm['pitch'] < 90.0 ) & good_mask )
aft_hrcs = ( (tlm['tscpos'] < -85000.0 ) & ( tlm['pitch'] > 90.0 ) & good_mask )
aft_hrci = ( ( -85000.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 0.0 ) & ( tlm['pitch'] > 90.0 ) & good_mask )
aft_aciss = ( ( 0.0 < tlm['tscpos'] ) & ( tlm['tscpos'] < 80000.0 ) & ( tlm['pitch'] > 90.0 ) & good_mask )
aft_acisi = ( (tlm['tscpos'] > 80000.0 ) & ( tlm['pitch'] > 90.0 ) & good_mask )
msid='1pdeaat'
plt.plot(pred[msid][fwd_hrci], tlm[msid][fwd_hrci] - pred[msid][fwd_hrci], "ob", markersize=5)
plt.plot(pred[msid][fwd_hrcs], tlm[msid][fwd_hrcs] - pred[msid][fwd_hrcs], "ok", markersize=5)
plt.plot(pred[msid][fwd_aciss], tlm[msid][fwd_aciss] - pred[msid][fwd_aciss], "or", markersize=5)
plt.plot(pred[msid][fwd_acisi], tlm[msid][fwd_acisi] - pred[msid][fwd_acisi], "og", markersize=5)
plt.plot(pred[msid][m80_hrci], tlm[msid][m80_hrci] - pred[msid][m80_hrci], "vb", markersize=5)
plt.plot(pred[msid][m80_hrcs], tlm[msid][m80_hrcs] - pred[msid][m80_hrcs], "vk", markersize=5)
plt.plot(pred[msid][m80_aciss], tlm[msid][m80_aciss] - pred[msid][m80_aciss], "vr", markersize=5)
plt.plot(pred[msid][m80_acisi], tlm[msid][m80_acisi] - pred[msid][m80_acisi], "vg", markersize=5)
plt.plot(pred[msid][mid_hrci], tlm[msid][mid_hrci] - pred[msid][mid_hrci], "^b", markersize=5)
plt.plot(pred[msid][mid_hrcs], tlm[msid][mid_hrcs] - pred[msid][mid_hrcs], "^k", markersize=5)
plt.plot(pred[msid][mid_aciss], tlm[msid][mid_aciss] - pred[msid][mid_aciss], "^r", markersize=5)
plt.plot(pred[msid][mid_acisi], tlm[msid][mid_acisi] - pred[msid][mid_acisi], "^g", markersize=5)
plt.plot(pred[msid][aft_hrci], tlm[msid][aft_hrci] - pred[msid][aft_hrci], ".b", markersize=2)
plt.plot(pred[msid][aft_hrcs], tlm[msid][aft_hrcs] - pred[msid][aft_hrcs], ".k", markersize=2)
plt.plot(pred[msid][aft_aciss], tlm[msid][aft_aciss] - pred[msid][aft_aciss], ".r", markersize=2)
plt.plot(pred[msid][aft_acisi], tlm[msid][aft_acisi] - pred[msid][aft_acisi], ".g", markersize=2)
maxmodeltemp=ndarray.max(pred[msid][good_mask])
maxresid=ndarray.max(tlm[msid][good_mask]-pred[msid][good_mask])
x = np.array(np.linspace(52.5-maxresid,maxmodeltemp,num=5))
my_y = 52.5 - x
plt.plot( x, my_y )
plt.ylabel('Data - Model')
plt.xlabel('1pdeaat Model')
plt.title('blue,black,red,green=hrci,hrcs,aciss,acisi, 45<o<65<v<80<^<90<.')
plt.grid()
# raise ValueError
outfile=os.path.join(outdir,'1pdeaat_resid.png')
fig.savefig(outfile)
return plots
def main_test():
# As a main function, this function tests other functions written in domain_functions.py:
# - compute triangulation of domain
# -
nodes = np.array([[0., 0.], [0., 10.], [15., 10.], [15., 0.], [6., 10.],
[9., 10.], [11., 0.], [2., 8.], [8., 7.], [11., 2.],
[5., 5.]])
tri = spatial.Delaunay(nodes)
print ''
print 'Entered points, triangles by vertex, and neighbors:'
print tri.points # Numbering of the points is in order of input
print ''
print tri.simplices # From point numbers, give the triangles, starting with # 0
print ''
print tri.neighbors # Give the triangle number of neighbors, and -1 for boundary
# Test which triangle:
# print tri.find_simplex(np.array([2., 2.]))
# Choose a simplex and facet control idx:
num = 0
facet_id = 0
# Extract points:
points = tri.points
vertices = tri.simplices
vertices_here = vertices[num, :]
print ''
print 'vertices_here:'
print vertices_here
# Extract normals:
nhat_array = nhat_tri_facet(tri, num)
print ''
print 'Local normal vectors:'
print nhat_array
u_optimal, fg_array = u_for_F_v1(tri, num, 0.01, np.eye(2), facet_id)
u_stay, fg_array_stay = u_for_stay_v1(tri, num, 0.01, np.eye(2))
print ''
print 'Optimal u at each vertex:'
print u_optimal
print ''
print 'fg_array for select triangle and facet:'
print fg_array
print ''
print 'Optimal u to stay:'
print u_stay
fg_master_array = fg_for_domain(tri, 0.01, np.eye(2), 2)
print ''
print 'fg_master_array:'
print fg_master_array
x_array = points[vertices_here, 0]
y_array = points[vertices_here, 1]
x_lim = np.array([ndarray.min(x_array), ndarray.max(x_array)])
y_lim = np.array([ndarray.min(y_array), ndarray.max(y_array)])
domain_plot = 1
control_plot = 1
# Plot partitioning of domain:
if domain_plot == 1:
plt.rc('text', usetex=True)
plt.rc('font', family='serif')
plt.triplot(nodes[:, 0], nodes[:, 1], tri.simplices.copy())
plt.plot(nodes[:, 0], nodes[:, 1], 'o')
plt.title('Partitioned Domain')
plt.grid(linestyle="--", linewidth=0.1, color='.25', zorder=-10)
# Quiver velocity field to leave selected triangle:
if control_plot == 1:
X, Y = np.meshgrid(np.arange(x_lim[0] - 0.5, x_lim[1] + 0.5, 0.5),
np.arange(y_lim[0] - 0.5, y_lim[1] + 0.5, 0.5))
U = fg_array[0, 0] * X + fg_array[0, 1] * Y + fg_array[0, 2]
V = fg_array[1, 0] * X + fg_array[1, 1] * Y + fg_array[1, 2]
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
plt.rc('text', usetex=True)
plt.rc('font', family='serif')
plt.title(
'Control Vector Field for Chosen Simplex - Exit Chosen Facet')
ax.triplot(nodes[:, 0], nodes[:, 1], tri.simplices.copy())
ax.plot(nodes[:, 0], nodes[:, 1], 'o')
ax.quiver(X, Y, U, V)
ax.grid(linestyle="--", linewidth=0.1, color='.25', zorder=-10)
# ax.set_xlim(x_lim)
# ax.set_ylim(y_lim)
X, Y = np.meshgrid(np.arange(x_lim[0] - 0.5, x_lim[1] + 0.5, 0.5),
np.arange(y_lim[0] - 0.5, y_lim[1] + 0.5, 0.5))
U = fg_array_stay[0, 0] * X + fg_array_stay[0, 1] * Y + fg_array_stay[
0, 2]
V = fg_array_stay[1, 0] * X + fg_array_stay[1, 1] * Y + fg_array_stay[
1, 2]
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
plt.rc('text', usetex=True)
plt.rc('font', family='serif')
plt.title('Control Vector Field for Chosen Simplex - Remain')
ax.triplot(nodes[:, 0], nodes[:, 1], tri.simplices.copy())
ax.plot(nodes[:, 0], nodes[:, 1], 'o')
ax.quiver(X, Y, U, V)
ax.grid(linestyle="--", linewidth=0.1, color='.25', zorder=-10)
# ax.set_xlim(x_lim)
# ax.set_ylim(y_lim)
if (domain_plot == 1) or (control_plot == 1):
plt.show()
return
# Execute test:
# main_test()
def main_func():
image = cv2.imread('asl_alphabet.jpg')
resized_image = cv2.resize(image, (570, 720))
# loading the pre-trained model
model = load_model('SeqModel.h5')
camera = cv2.VideoCapture(0) # 0 -> index of camera
camera.set(3, 1080)
camera.set(4, 720)
predict_letter_list = []
word_str = ''
word_list = []
while True:
check, frame1 = camera.read()
cv2.rectangle(frame1, (20, 70), (275, 325), (0, 255, 0), thickness=1)
gray_frame = cv2.cvtColor(frame1, cv2.COLOR_BGR2GRAY)
median_blurred = cv2.medianBlur(gray_frame,
3) # blurring (median) the gray frame
gaus_blurred = cv2.GaussianBlur(
median_blurred, (3, 3),
0) # blurring (gaussian) the already blurred gray frame
canny_frame = cv2.Canny(gaus_blurred, 30, 30)
cropped_frame = canny_frame[71:325, 21:275]
cropFrame_array = feed_preprocessing(cropped_frame)
# predict
predictions_prob = model.predict(cropFrame_array)
# print(predictions_prob)
# take the value with the highest probability
most_prob = ndarray.max(predictions_prob)
# print(most_prob)
# [0] -> A , [1] -> B , [2]-> C
# print(type(predictions_label)) --> numpy.ndarray
predictions_label = model.predict_classes(cropFrame_array)
# convert [0] to 0 for future operations
string_label = functools.reduce(lambda x, y: x + str(y),
predictions_label, '')
integer_label = int(string_label)
# 0 -> A, 1 -> B, 2-> C
# print(integer_label)
# matching the labels to the actual values
# most_prob > number (where number determined based on measures) to avoid random results
if most_prob > 0.65:
# print(f'accuracy: {most_prob}')
# print(f'letter {write_letter(integer_label)}')
letter = write_letter(integer_label)
# append the predicted letter into the list
predict_letter_list.append(letter)
print(predict_letter_list)
# every x letters
if len(predict_letter_list) % 15 == 0:
# check that ALL (x) the letters in the list are the same
if all(x == predict_letter_list[0]
for x in predict_letter_list):
if integer_label == 20:
word_list += ' '
elif integer_label == 4: # trycatch should be added for empty list(deleting letter from empty word)
try:
word_list.pop()
except:
pass
else:
# add to the word that will be printed one of the identical items of the list
word_list.append(predict_letter_list[0])
time.sleep(.5)
print(word_list)
# every x letter clear the list so the process repeats again from the start
predict_letter_list.clear()
# convert list into string for putText function
word_str = "".join(word_list)
frame1 = cv2.putText(frame1, 'Predicted Word: ' + word_str, (20, 60),
cv2.FONT_HERSHEY_SIMPLEX, 2, (128, 0, 0), 3)
# frame1 = cv2.putText(frame1, 'Predicted Letter: ' + write_letter(integer_label), (20, 500),
# cv2.FONT_HERSHEY_SIMPLEX, 2, (128, 0, 0), 3)
if not check:
break
cv2.imshow('WINDOW', frame1)
cv2.imshow('CROPPED', cropped_frame)
cv2.imshow('ALPHABET', resized_image)
if frame1 is None:
break
keyboard = cv2.waitKey(1) & 0xff
if keyboard == 27: # esc to terminate
break
camera.release()
cv2.destroyAllWindows()