def recognize_from_video(video):
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
detector = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
if video == '0':
print('[INFO] Webcam mode is activated')
capture = cv2.VideoCapture(0)
if not capture.isOpened():
print("[ERROR] webcamera not found")
sys.exit(1)
else:
if pathlib.Path(video).exists():
capture = cv2.VideoCapture(video)
while(True):
ret, img = capture.read()
objs = detect_objects(img, detector)
for obj in objs:
dbface_utils.drawbbox(img, obj)
cv2.imshow('frame', img)
# press q to end video capture
if cv2.waitKey(1)&0xFF == ord('q'):
break
if not ret:
continue
capture.release()
cv2.destroyAllWindows()
print('Script finished successfully.')
def recognize_from_image():
# prepare input data
img = load_image(args.input)
print(f'input image shape: {img.shape}')
# net initialize
detector = ailia.Detector(
MODEL_PATH,
WEIGHT_PATH,
len(COCO_CATEGORY),
format=ailia.NETWORK_IMAGE_FORMAT_RGB,
channel=ailia.NETWORK_IMAGE_CHANNEL_FIRST,
range=ailia.NETWORK_IMAGE_RANGE_U_FP32,
algorithm=ailia.DETECTOR_ALGORITHM_YOLOV3,
env_id=args.env_id,
)
pose = ailia.Net(POSE_MODEL_PATH, POSE_WEIGHT_PATH, env_id=args.env_id)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
detector.compute(img, THRESHOLD, IOU)
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
detector.compute(img, THRESHOLD, IOU)
# plot result
res_img = plot_results(detector, pose, img, COCO_CATEGORY)
cv2.imwrite(args.savepath, res_img)
print('Script finished successfully.')
def recognize_from_image():
# prepare input data
org_img = load_image(
args.input,
(IMAGE_HEIGHT, IMAGE_WIDTH),
)
input_data = load_image(args.input, (IMAGE_HEIGHT, IMAGE_WIDTH),
normalize_type='127.5',
gen_input_ailia=True)
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
preds_ailia = net.predict([input_data])
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
preds_ailia = net.predict([input_data])
# postprocessing
detections = but.postprocess(preds_ailia)
# generate detections
for detection in detections:
but.plot_detections(org_img, detection, save_image_path=args.savepath)
print('Script finished successfully.')
def main():
info = {
("paris-streetview", "rect"):
(WEIGHT_PARIS_STREETVIEW_PATH, MODEL_PARIS_STREETVIEW_PATH, (256,
256)),
("celebahq", "rect"):
(WEIGHT_CELEBAHQ_256_PATH, MODEL_CELEBAHQ_256_PATH, (256, 256)),
("celebahq-512", "rect"):
(WEIGHT_CELEBAHQ_512_PATH, MODEL_CELEBAHQ_512_PATH, (512, 512)),
("celebahq-512", "stroke"): (WEIGHT_CELEBAHQ_FREEFORM_PATH,
MODEL_CELEBAHQ_FREEFORM_PATH, (512, 512)),
("places2", "stroke"):
(WEIGHT_PLACE2_PATH, MODEL_PLACE2_PATH, (512, 680)),
}
key = (args.model, args.mask_type)
if key not in info:
logger.error("(MODEL = %s, MASK_TYPE = %s) is unmatch." % key)
logger.info("appropriate settings:\n"
"\t(MODEL = paris-streetview, MASK_TYPE = rect)\n"
"\t(MODEL = celebahq, MASK_TYPE = rect)\n"
"\t(MODEL = celebahq-512, MASK_TYPE = rect)\n"
"\t(MODEL = celebahq-512, MASK_TYPE = stroke)\n"
"\t(MODEL = places2, MASK_TYPE = stroke)")
sys.exit(-1)
# model files check and download
weight_path, model_path, img_shape = info[key]
check_and_download_models(weight_path, model_path, REMOTE_PATH)
# net initialize
net = ailia.Net(model_path, weight_path, env_id=args.env_id)
recognize_from_image(net, img_shape)
def main():
# model files check and download
check_and_download_models(WEIGHT_PATH, MODEL_PATH, REMOTE_PATH)
# load audio
for input_data_path in args.input:
logger.info('=' * 80)
logger.info(f'input: {input_data_path}')
data = sf.read(input_data_path)
# create instance
session = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
# inference
logger.info('Start inference...')
if args.benchmark:
logger.info('BENCHMARK mode')
for c in range(5):
start = int(round(time.time() * 1000))
label, conf = crnn(data, session)
end = int(round(time.time() * 1000))
logger.info("\tailia processing time {} ms".format(end -
start))
else:
label, conf = crnn(data, session)
logger.info(label)
logger.info(conf)
logger.info('Script finished successfully.')
def enhance_image():
for image_path in args.input:
# prepare input data
img = cv2.imread(image_path, cv2.IMREAD_UNCHANGED)
img = cv2.resize(img, dsize=(H, W))
# net initialize
model = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
upsampler = RealESRGAN(model)
# inference
logger.info('Start inference...')
if args.benchmark:
logger.info('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
output = upsampler.enhance(img)
end = int(round(time.time() * 1000))
logger.info(f'\tailia processing time {end - start} ms')
else:
output = upsampler.enhance(img)
savepath = get_savepath(args.savepath, image_path)
logger.info(f'saved at : {savepath}')
cv2.imwrite(savepath, output)
logger.info('Script finished successfully.')
def recognize_from_image():
# prepare input data
input_data = load_image(
args.input,
(IMAGE_HEIGHT, IMAGE_WIDTH),
normalize_type='ImageNet',
gen_input_ailia=True
)
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
preds_ailia = net.predict(input_data)
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
preds_ailia = net.predict(input_data)
# postprocessing
print_results(preds_ailia, vgg16_labels.imagenet_category)
print('Script finished successfully.')
def process_video():
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
if args.face_recognition:
locator = FaceLocator()
else:
locator = None
if args.video == '0':
print('[INFO] Webcam mode is activated')
capture = cv2.VideoCapture(0)
if not capture.isOpened():
print("[ERROR] webcamera not found")
sys.exit(1)
else:
if check_file_existance(args.video):
capture = cv2.VideoCapture(args.video)
while (True):
ret, frame = capture.read()
if cv2.waitKey(1) & 0xFF == ord('q'):
break
if not ret:
continue
img = process_frame(net, locator, frame)
cv2.imshow('frame', img[..., ::-1])
capture.release()
cv2.destroyAllWindows()
print('Script finished successfully.')
def recognize_from_image():
# prepare input data
input_img = cv2.imread(args.input)
data = load_image(args.input, (IMAGE_HEIGHT, IMAGE_WIDTH),
normalize_type='255',
gen_input_ailia=True)
# net initalize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
# compute execution time
for i in range(5):
start = int(round(time.time() * 1000))
preds_ailia = net.predict(data)[0]
end = int(round(time.time() * 1000))
print(f'ailia processing time {end - start} ms')
visualize_plots(input_img, preds_ailia)
cv2.imwrite(args.savepath, input_img)
# Confidence Map
channels = preds_ailia.shape[0]
cols = 8
plot_images('confidence',
preds_ailia,
tile_shape=((int)((channels + cols - 1) / cols), cols))
print('Script finished successfully.')
def wavfile_input_recognition():
if args.beamdecode:
try:
from ctcdecode import CTCBeamDecoder
except ImportError:
raise ImportError("BeamCTCDecoder requires paddledecoder package.")
decoder = CTCBeamDecoder(
LABELS,
LM_PATH,
ALPHA,
BETA,
CUTOFF_TOP_N,
CUTOFF_PROB,
BEAM_WIDTH,
NUM_PROCESS,
BRANK_LABEL_INDEX,
)
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
for soundf_path in args.input:
logger.info(soundf_path)
if args.ailia_audio:
wav,sr = sf.read(soundf_path)
wav = ailia.audio.resample(wav,sr,SAMPLING_RATE)
else:
wav = librosa.load(soundf_path, sr=SAMPLING_RATE)[0]
spectrogram = create_spectrogram(wav)
net.set_input_shape(spectrogram[0].shape)
# inference
logger.info('Start inference...')
if args.benchmark:
logger.info('BENCHMARK mode')
for c in range(5):
start = int(round(time.time() * 1000))
preds_ailia, output_length = net.predict(spectrogram)
end = int(round(time.time() * 1000))
logger.info("\tailia processing time {} ms".format(end-start))
else:
# Deep Speech output: output_probability, output_length
preds_ailia, output_length = net.predict(spectrogram)
if args.beamdecode:
text = beam_ctc_decode(
torch.from_numpy(preds_ailia),
torch.from_numpy(output_length),
decoder,
)
else:
text = decode(preds_ailia[0], output_length)
savepath = get_savepath(args.savepath, soundf_path, ext='.txt')
logger.info(f'Results saved at : {savepath}')
with open(savepath, 'w', encoding='utf-8') as f:
f.write(text)
logger.info(f'predict sentence:\n{text}')
logger.info('Script finished successfully.')
def transform_image():
"""Full transormation on a single image loaded from filepath in arguments."""
image = cv2.imread(args.input)
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
if args.face_recognition:
locator = FaceLocator()
else:
locator = None
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
out_image = process_frame(net, locator, image)
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
out_image = process_frame(net, locator, image)
cv2.imwrite(args.savepath, out_image[..., ::-1])
return True
def recognize_from_image():
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
# input image loop
for image_path in args.input:
# prepare input data
logger.info(image_path)
input_img = load_image(
image_path,
(IMAGE_HEIGHT, IMAGE_WIDTH),
normalize_type='None',
)
input_data = get_processed_image(input_img)
# inference
logger.info('Start inference...')
if args.benchmark:
logger.info('BENCHMARK mode')
for i in range(args.benchmark_count):
start = int(round(time.time() * 1000))
preds = net.predict(input_data)
end = int(round(time.time() * 1000))
logger.info(f'\tailia processing time {end - start} ms')
else:
preds = net.predict(input_data)
# show results
print_results(preds, inceptionv4_labels.imagenet_category)
logger.info('Script finished successfully.')
def main():
# model files check and download
check_and_download_models(WEIGHT_PATH, MODEL_PATH, REMOTE_PATH)
if args.arch == 'bert-base-cased' or args.arch == 'bert-base-uncased':
tokenizer = BertTokenizer.from_pretrained(args.arch)
else:
tokenizer = BertJapaneseTokenizer.from_pretrained("cl-tohoku/" +
args.arch)
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
net.set_input_blob_shape((1, PADDING_LEN),
net.find_blob_index_by_name("token_type_ids"))
net.set_input_blob_shape((1, PADDING_LEN),
net.find_blob_index_by_name("input_ids"))
net.set_input_blob_shape((1, PADDING_LEN),
net.find_blob_index_by_name("attention_mask"))
with codecs.open(args.input[0], 'r', 'utf-8', 'ignore') as f:
s = f.readlines()
for text in s:
tokenized_text = tokenizer.tokenize(text)
original_text_len = len(tokenized_text)
for j in range(len(tokenized_text), PADDING_LEN):
tokenized_text.append('[PAD]')
score = numpy.zeros((len(tokenized_text)))
suggest = {}
for i in range(0, len(tokenized_text)):
masked_index = i
if tokenized_text[masked_index] == '[PAD]':
continue
tokenized_text_saved = tokenized_text[masked_index]
tokenized_text[masked_index] = '[MASK]'
outputs = inference(net, tokenizer, tokenized_text, masked_index,
original_text_len)
target_ids = tokenizer.convert_tokens_to_ids(
[tokenized_text_saved])
index = target_ids[0]
score[masked_index] = outputs[0][0, masked_index][index]
predictions = torch.from_numpy(outputs[0][0, masked_index]).topk(1)
index = predictions.indices[0]
top_token = tokenizer.convert_ids_to_tokens([index])[0]
suggest[masked_index] = top_token
tokenized_text[masked_index] = tokenized_text_saved
fine_text = colorize(tokenized_text, score, suggest)
print(fine_text)
print('Script finished successfully.')
def recognize_from_image():
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
# input image loop
for image_path in args.input:
# prepare input data
logger.info(image_path)
src_img = cv2.imread(image_path)
input_data = load_image(
image_path,
(IMAGE_HEIGHT, IMAGE_WIDTH),
)
input_data = input_data[np.newaxis, :, :, :]
net.set_input_shape(input_data.shape)
# inference
logger.info('Start inference...')
if args.benchmark:
logger.info('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
preds_ailia = net.predict(input_data)
end = int(round(time.time() * 1000))
logger.info(f'\tailia processing time {end - start} ms')
else:
preds_ailia = net.predict(input_data)
# postprocessing
pred = preds_ailia.reshape((IMAGE_HEIGHT, IMAGE_WIDTH))
dst = transfer(src_img, pred)
savepath = get_savepath(args.savepath, image_path)
logger.info(f'saved at : {savepath}')
cv2.imwrite(savepath, dst)
logger.info('Script finished successfully.')
def recognize_from_video():
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
if args.video == '0':
print('[INFO] Webcam mode is activated')
capture = cv2.VideoCapture(0)
if not capture.isOpened():
print("[ERROR] webcamera not found")
sys.exit(1)
else:
if check_file_existance(args.video):
capture = cv2.VideoCapture(args.video)
while (True):
ret, frame = capture.read()
if cv2.waitKey(1) & 0xFF == ord('q'):
break
if not ret:
continue
frame = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
input_data = preprocess(frame)
net.set_input_shape(input_data.shape)
boxes, labels, scores, masks = net.predict([input_data])
display_objdetect_image(frame, boxes, labels, scores, masks)
plt.pause(.01)
capture.release()
cv2.destroyAllWindows()
print('Script finished successfully.')
def recognize_from_image():
# prepare input data
image = Image.open(args.input)
input_data = preprocess(image)
# net initialize
# This model requires fuge gpu memory so fallback to cpu mode
env_id = args.env_id
if env_id != -1 and ailia.get_environment(env_id).props == "LOWPOWER":
env_id = -1
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
net.set_input_shape(input_data.shape)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
boxes, labels, scores, masks = net.predict([input_data])
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
boxes, labels, scores, masks = net.predict([input_data])
# postprocessing
fig, ax = create_figure()
display_objdetect_image(
fig, ax, image, boxes, labels, scores, masks, savepath=args.savepath
)
print('Script finished successfully.')
def main():
# model files check and download
info = {
'lip': (WEIGHT_LIP_PATH, MODEL_LIP_PATH,
(IMAGE_LIP_SIZE, IMAGE_LIP_SIZE), CATEGORY_LIP),
'atr': (WEIGHT_ATR_PATH, MODEL_ATR_PATH,
(IMAGE_ATR_SIZE, IMAGE_ATR_SIZE), CATEGORY_ATR),
'pascal': (WEIGHT_PASCAL_PATH, MODEL_PASCAL_PATH,
(IMAGE_PASCAL_SIZE, IMAGE_ATR_SIZE), CATEGORY_PASCAL),
}
weight_path, model_path, img_size, category = info[args.arch]
check_and_download_models(weight_path, model_path, REMOTE_PATH)
# Workaround for accuracy issue on
# ailia SDK 1.2.4 + opset11 + gpu (metal/vulkan)
detector = ailia.Net(model_path, weight_path, env_id=args.env_id)
params = {'img_size': img_size, 'category': category}
if args.video is not None:
# video mode
recognize_from_video(args.video, detector, params)
else:
# image mode
# input image loop
for image_path in args.input:
# prepare input data
logger.info(image_path)
recognize_from_image(image_path, detector, params)
logger.info('Script finished successfully.')
def recognize_from_image():
# prepare input data
input_data = load_image(args.input, (IMAGE_HEIGHT, IMAGE_WIDTH),
gen_input_ailia=True)
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
preds_ailia = net.predict(input_data)
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
preds_ailia = net.predict(input_data)
# postprocessing
if args.smooth:
preds_ailia = smooth_output(preds_ailia)
save_pred(preds_ailia, args.savepath, IMAGE_HEIGHT, IMAGE_WIDTH)
print('Script finished successfully.')
def enhance_video():
# net initialize
model = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
upsampler = RealESRGAN(model)
capture = get_capture(args.video)
# create video writer if savepath is specified as video format
if args.savepath != SAVE_IMAGE_PATH:
logger.warning(
'currently, video results cannot be output correctly...')
f_h = int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT))
f_w = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH))
save_h, save_w = calc_adjust_fsize(f_h, f_w, IMAGE_HEIGHT, IMAGE_WIDTH)
writer = get_writer(args.savepath, save_h, save_w * 2)
else:
writer = None
while (True):
ret, frame = capture.read()
if (cv2.waitKey(1) & 0xFF == ord('q')) or not ret:
break
img = cv2.resize(frame, dsize=(H, W))
# inference
output = upsampler.enhance(img)
#plot result
cv2.imshow('frame', output)
if writer is not None:
writer.release()
logger.info('Script finished successfully.')
def microphone_input_recognition():
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
if args.beamdecode:
try:
from ctcdecode import CTCBeamDecoder
except ImportError:
raise ImportError("BeamCTCDecoder requires paddledecoder package.")
decoder = CTCBeamDecoder(LABELS, LM_PATH, ALPHA, BETA, CUTOFF_TOP_N, CUTOFF_PROB, BEAM_WIDTH,
NUM_PROCESS, BRANK_LABEL_INDEX)
while True:
wav = record_microphone_input()
spectrogram = create_spectrogram(wav)
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
net.set_input_shape(spectrogram[0].shape)
# inference
print('Translating...')
#Deep Speech output: output_probability, output_length
preds_ailia, output_length = net.predict(spectrogram)
if args.beamdecode:
text = beam_ctc_decode(torch.from_numpy(preds_ailia), torch.from_numpy(output_length), decoder)
else:
text = decode(preds_ailia[0], output_length)
print(f'predict sentence:\n{text}\n')
time.sleep(1)
def extract_feature_vec_from_image():
# prepare input data
input_img = load_image(
args.input,
(IMAGE_HEIGHT, IMAGE_WIDTH),
normalize_type='None',
)
input_data = prepare_input_data(input_img)
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
fe_net = ailia.Net(FE_MODEL_PATH, FE_WEIGHT_PATH, env_id=env_id)
fe_net.set_input_shape(input_data.shape)
input_dict = {'data': input_data}
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
_ = fe_net.predict(input_dict)[0]
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
_ = fe_net.predict(input_dict)[0]
# Extracting the output of a specifc layer
idx = fe_net.find_blob_index_by_name('encode1')
preds_ailia = fe_net.get_blob_data(idx)
print(preds_ailia.reshape(preds_ailia.shape[0], -1))
print('Script finished successfully.')
def recognize_from_image():
# prepare input data
input_data = load_image(args.input, (IMAGE_HEIGHT, IMAGE_WIDTH),
normalize_type='ImageNet',
gen_input_ailia=True)
src_img = load_image(args.input, (IMAGE_HEIGHT, IMAGE_WIDTH),
normalize_type='None')
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
preds_ailia = net.predict(input_data)
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
preds_ailia = net.predict(input_data)
# postprocessing
res_img = postprocess(src_img, preds_ailia)
cv2.imwrite(args.savepath, res_img)
print('Script finished successfully.')
def recognize_from_image():
# prepare input data
input_data = load_image(args.input, (IMAGE_HEIGHT, IMAGE_WIDTH),
normalize_type='255',
gen_input_ailia=True)
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
preds_ailia = net.predict(input_data)
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
preds_ailia = net.predict(input_data)
# postprocessing
output_img = preds_ailia[0].transpose((1, 2, 0))
output_img = cv2.cvtColor(output_img, cv2.COLOR_RGB2BGR)
cv2.imwrite(args.savepath, output_img * 255)
print('Script finished successfully.')
def main():
# model files check and download
check_and_download_models(WEIGHT_PATH, MODEL_PATH, REMOTE_PATH)
ailia_model = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
tokenizer = DistilBertTokenizer.from_pretrained(
'distilbert-base-uncased-finetuned-sst-2-english')
model_inputs = tokenizer.encode_plus(args.input, return_tensors="pt")
inputs_onnx = {
k: v.cpu().detach().numpy()
for k, v in model_inputs.items()
}
logger.info("Input : " + str(args.input))
# inference
if args.benchmark:
logger.info('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
score = ailia_model.predict(inputs_onnx)
end = int(round(time.time() * 1000))
logger.info("\tailia processing time {} ms".format(end - start))
else:
score = ailia_model.predict(inputs_onnx)
score = numpy.exp(score) / numpy.exp(score).sum(-1, keepdims=True)
label_name = ["negative", "positive"]
label_id = numpy.argmax(numpy.array(score))
logger.info("Label : " + str(label_name[label_id]))
logger.info("Score : " + str(score[0][0][label_id]))
logger.info('Script finished successfully.')
def main():
dic_model = {
'blouse': (WEIGHT_BLOUSE_PATH, MODEL_BLOUSE_PATH, IMAGE_BLOUSE_PATH),
'dress': (WEIGHT_DRESS_PATH, MODEL_DRESS_PATH, IMAGE_DRESS_PATH),
'outwear':
(WEIGHT_OUTWEAR_PATH, MODEL_OUTWEAR_PATH, IMAGE_OUTWEAR_PATH),
'skirt': (WEIGHT_SKIRT_PATH, MODEL_SKIRT_PATH, IMAGE_SKIRT_PATH),
'trousers':
(WEIGHT_TROUSERS_PATH, MODEL_TROUSERS_PATH, IMAGE_TROUSERS_PATH),
}
weight_path, model_path, img_path = dic_model[args.clothing_type]
# model files check and download
check_and_download_models(weight_path, model_path, REMOTE_PATH)
# initialize
net = ailia.Net(model_path, weight_path, env_id=args.env_id)
if args.video is not None:
# video mode
recognize_from_video(args.video, net)
else:
# image mode
# input image loop
for image_path in args.input:
logger.info(image_path)
recognize_from_image(image_path, net)
logger.info('Script finished successfully.')
def main():
info = {
("celeba", 256):
(WEIGHT_CELEBA256_PATH, MODEL_CELEBA256_PATH, (256, 256)),
("places", 256):
(WEIGHT_PLACES256_PATH, MODEL_PLACES256_PATH, (256, 256)),
("places", 512):
(WEIGHT_PLACES512_PATH, MODEL_PLACES512_PATH, (512, 512)),
("places", 1024):
(WEIGHT_PLACES1024_PATH, MODEL_PLACES1024_PATH, (1024, 1024))
}
key = (args.model, args.img_res)
if key not in info:
logger.error("(MODEL = %s, IMG_RESOLUTION = %s) is unmatch." % key)
logger.info("appropriate settings:\n"
"\t(MODEL = celeba, IMG_RESOLUTION = 256)\n"
"\t(MODEL = places, IMG_RESOLUTION = 256 or 512 or 1024)")
sys.exit(-1)
if "FP16" in ailia.get_environment(
args.env_id).props or platform.system() == 'Darwin':
logger.warning('This model do not work on FP16. So use CPU mode.')
args.env_id = 0
# model files check and download
weight_path, model_path, img_shape = info[key]
check_and_download_models(weight_path, model_path, REMOTE_PATH)
# net initialize
net = ailia.Net(model_path, weight_path, env_id=args.env_id)
recognize_from_image(net, img_shape)
def compare_images():
# prepare input data
imgs_1 = prepare_input_data(args.inputs[0])
imgs_2 = prepare_input_data(args.inputs[1])
imgs = np.concatenate([imgs_1, imgs_2], axis=0)
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
# compute execution time
print('Start inference...')
for i in range(5):
start = int(round(time.time() * 1000))
preds_ailia = net.predict(imgs)
end = int(round(time.time() * 1000))
print(f'ailia processing time {end - start} ms')
# postprocessing
fe_1 = np.concatenate([preds_ailia[0], preds_ailia[1]], axis=0)
fe_2 = np.concatenate([preds_ailia[2], preds_ailia[3]], axis=0)
sim = cosin_metric(fe_1, fe_2)
print(f'Similarity of ({args.inputs[0]}, {args.inputs[1]}) : {sim:.3f}')
if THRESHOLD > sim:
print('They are not the same face!')
else:
print('They are the same face!')
def recognize_from_image():
# prepare input data
image = Image.open(args.input)
input_data = preprocess(image)
# net initialize
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
net.set_input_shape(input_data.shape)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
boxes, labels, scores, masks = net.predict([input_data])
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
boxes, labels, scores, masks = net.predict([input_data])
# postprocessing
display_objdetect_image(image,
boxes,
labels,
scores,
masks,
savepath=args.savepath)
print('Script finished successfully.')
def main():
# model files check and download
print("=== ST-GCN model ===")
check_and_download_models(WEIGHT_PATH, MODEL_PATH, REMOTE_PATH)
print("=== OpenPose model ===")
check_and_download_models(WEIGHT_POSE_PATH, MODEL_POSE_PATH,
REMOTE_POSE_PATH)
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
if args.arch == "pyopenpose":
pose = op.WrapperPython()
params = dict(model_folder='.', model_pose='COCO')
pose.configure(params)
pose.start()
else:
pose = ailia.PoseEstimator(MODEL_POSE_PATH,
WEIGHT_POSE_PATH,
env_id=args.env_id,
algorithm=POSE_ALGORITHM)
if args.arch == "openpose":
pose.set_threshold(0.1)
if args.video is not None:
# realtime mode
recognize_realtime(args.video, pose, net)
else:
# offline mode
recognize_from_file(args.input, pose, net)
def recognize_from_image(filename):
# load input image
img = load_image(filename)
print(f'input image shape: {img.shape}')
img = cv2.cvtColor(img, cv2.COLOR_BGRA2BGR)
env_id = ailia.get_gpu_environment_id()
print(f'env_id: {env_id}')
detector = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=env_id)
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for i in range(5):
start = int(round(time.time() * 1000))
objs = detect_objects(img, detector)
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
objs = detect_objects(img, detector)
# show image
for obj in objs:
dbface_utils.drawbbox(img, obj)
cv2.imwrite(args.savepath, img)
print('Script finished successfully.')
