def camera_predict():
video_captor = cv2.VideoCapture(0)
predicted_class = None
while True:
ret, frame = video_captor.read()
face_img, face_coor = camera_face_detect.face_d(frame)
if face_coor is not None:
[x_screen, y_screen, w_screen, h_screen] = face_coor
cv2.rectangle(frame, (x_screen, y_screen),
(x_screen + w_screen, y_screen + h_screen),
(255, 0, 0), 2)
if cv2.waitKey(1) & 0xFF == ord(' '):
face_img, face_coor = camera_face_detect.face_d(frame)
if face_coor is not None:
[x, y, w, h] = face_coor
cv2.rectangle(frame, (x, y), (x + w, y + h), (255, 0, 0), 2)
if face_img is not None:
if not os.path.exists(cfg.output_folder):
os.mkdir('./camera_output')
cv2.imwrite(os.path.join(cfg.output_folder, 'face_image.jpg'),
face_img)
gray = cv2.cvtColor(face_img, cv2.COLOR_RGB2GRAY)
gray = cv2.resize(gray, (240, 240))
img = gray[:, :, np.newaxis]
img = np.concatenate((img, img, img), axis=2)
img = Image.fromarray(np.uint8(img))
inputs = transform_test(img)
class_names = [
'Angry', 'Disgusted', 'Fearful', 'Happy', 'Sad',
'Surprised', 'Neutral'
]
net = VGG16.Net()
checkpoint = torch.load(cfg.ckpt_path)
net.load_state_dict(checkpoint['net'])
if use_cuda:
net.to(device)
net.eval()
ncrops, c, h, w = np.shape(inputs)
inputs = inputs.view(-1, c, h, w)
inputs = Variable(inputs, volatile=True)
if use_cuda:
inputs = inputs.to(device)
outputs = net(inputs)
outputs_avg = outputs.view(ncrops, -1).mean(0)
score = F.softmax(outputs_avg)
print(score)
_, predicted = torch.max(outputs_avg.data, 0)
predicted_class = class_names[int(predicted.cpu().numpy())]
print(predicted_class)
if predicted_class is not None:
cv2.putText(frame, predicted_class, (30, 60),
cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 0, 255), 1)
cv2.imwrite(os.path.join(cfg.output_folder, 'predict.jpg'),
frame)
if predicted_class is not None:
cv2.putText(frame, predicted_class, (30, 60),
cv2.FONT_HERSHEY_SIMPLEX, 2, (0, 0, 255), 1)
cv2.imshow('camera', frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cfg.data_path, 'train'),
transform=transform_train)
test_data = torchvision.datasets.ImageFolder(os.path.join(
cfg.data_path, 'test'),
transform=transform_test)
train_loader = torch.utils.data.DataLoader(train_data,
batch_size=cfg.bs,
shuffle=True)
test_loader = torch.utils.data.DataLoader(test_data,
batch_size=cfg.bs,
shuffle=True)
net = VGG16.Net()
if cfg.resume:
print('------------------------------')
print('==> Loading the checkpoint ')
if not os.path.exists(cfg.ckpt_path):
raise AssertionError['Can not find path']
checkpoint = torch.load(cfg.ckpt_path)
net.load_state_dict(checkpoint['net'])
best_test_acc = checkpoint['best_test_acc']
print('best_test_acc is %.4f%%' % best_test_acc)
best_test_acc_epoch = checkpoint['best_test_acc_epoch']
print('best_test_acc_epoch is %d' % best_test_acc_epoch)
start_epoch = checkpoint['best_test_acc_epoch'] + 1
else:
print('------------------------------')
def image_predict(image):
image_path = os.path.join(arg.input_folder, image)
src_img = np.array(Image.open(image_path))
face_img, face_coor = image_face_detect.face_d(src_img)
gray = cv2.cvtColor(face_img, cv2.COLOR_RGB2GRAY)
gray = cv2.resize(gray, (240, 240))
img = gray[:, :, np.newaxis]
img = np.concatenate((img, img, img), axis=2)
img = Image.fromarray(np.uint8(img))
inputs = transform_test(img)
class_names = [
'Angry', 'Disgusted', 'Fearful', 'Happy', 'Sad', 'Surprised', 'Neutral'
]
net = VGG16.Net()
checkpoint = torch.load(arg.ckpt_path)
net.load_state_dict(checkpoint['net'])
if use_cuda:
net.cuda()
net.eval()
ncrops, c, h, w = np.shape(inputs)
inputs = inputs.view(-1, c, h, w)
inputs = Variable(inputs, volatile=True)
if use_cuda:
inputs = inputs.to(device)
outputs = net(inputs)
outputs_avg = outputs.view(ncrops, -1).mean(0)
score = F.softmax(outputs_avg)
_, predicted = torch.max(outputs_avg.data, 0)
expression = class_names[int(predicted.cpu().numpy())]
if face_coor is not None:
[x, y, w, h] = face_coor
cv2.rectangle(src_img, (x, y), (x + w, y + h), (255, 0, 0), 2)
plt.rcParams['figure.figsize'] = (11, 6)
axes = plt.subplot(1, 2, 1)
plt.imshow(src_img)
plt.title('Input Image', fontsize=20)
axes.set_xticks([])
axes.set_yticks([])
plt.tight_layout()
plt.subplots_adjust(left=0.05,
bottom=0.2,
right=0.95,
top=0.9,
hspace=0.02,
wspace=0.3)
plt.subplot(1, 2, 2)
ind = 0.1 + 0.6 * np.arange(len(class_names))
width = 0.4
for i in range(len(class_names)):
plt.bar(ind[i], score.data.cpu().numpy()[i], width, color='orangered')
plt.title("Result Analysis", fontsize=20)
plt.xticks(ind, class_names, rotation=30, fontsize=12)
if arg.SAVE_FLAG:
if not os.path.exists(arg.output_folder):
os.mkdir('./image_output')
save_path = os.path.join(arg.output_folder, image)
plt.savefig(save_path + '-result' + '.jpg')
else:
if arg.show_resultimg:
plt.show()
print("The Expression is %s" % expression)