def main(args):
helper = d.DemoHelper()
helper.parse_arguments(
args,
"Runs the given ELL model passing images from camera or static image file\n"
"Either the ELL model file, or the compiled model's Python module must be given,\n"
"using the --model or --compiledModel options respectively.\n"
"Example:\n"
" python demo.py categories.txt --compiledModel tutorial1/pi3/model1\n"
" python demo.py categories.txt --model model1.ell\n"
"This shows opencv window with image classified by the model using given labels"
)
# Initialize image source
helper.init_image_source()
lastPrediction = ""
while (not helper.done()):
# Grab next frame
frame = helper.get_next_frame()
# Prepare the image to send to the model.
# This involves scaling to the required input dimension and re-ordering from BGR to RGB
data = helper.prepare_image_for_predictor(frame)
# Get the model to classify the image, by returning a list of probabilities for the classes it can detect
predictions = helper.predict(data)
# Get the (at most) top 5 predictions that meet our threshold. This is returned as a list of tuples,
# each with the text label and the prediction score.
top5 = helper.get_top_n_predictions(predictions, 5)
# Turn the top5 into a text string to display
text = ", ".join([
"(" + str(int(element[1] * 100)) + "%) " +
helper.get_label(element[0]) for element in top5
])
save = False
if (text != lastPrediction):
print(text)
save = True
lastPrediction = text
# Draw the text on the frame
if not helper.nogui:
frameToShow = frame
helper.draw_label(frameToShow, text)
helper.draw_fps(frameToShow)
# Show the new frame
helper.show_image(frameToShow, save)
helper.report_times()
def validate_image(args, filename):
"""Evaluates a single image to get predictions"""
helper = d.DemoHelper()
helper.parse_arguments(
args,
helpString=
"Evaluates a series of images to get predictions and performance metrics"
)
# Initialize image source
helper.init_image_source()
result = {}
result["input"] = basename(filename)
_logger.info(result["input"])
while (not helper.done()):
# Grab next frame
frame = helper.get_next_frame()
# Prepare the image to send to the model.
# This involves scaling to the required input dimension and re-ordering from BGR to RGB
data = helper.prepare_image_for_predictor(frame)
# Get the model to classify the image, by returning a list of probabilities for the classes it can detect
predictions = helper.predict(data)
# Get the (at most) top 5 predictions that meet our threshold. This is returned as a list of tuples,
# each with the text label and the prediction score.
top5 = helper.get_top_n_predictions(predictions, 5)
text = "".join([
str(element[0]) + "(" + str(int(100 * element[1])) + "%) "
for element in top5
])
_logger.info("\tPrediction: " + text)
# Ensure that numbers are JSON serializable
result["predictions"] = [(element[0], np.float(element[1]))
for element in top5]
result["avg_time"] = helper.get_times()
helper.report_times(node_level=False)
return result
def main():
helper = d.DemoHelper()
arg_parser = argparse.ArgumentParser(
"Runs the given ELL model passing images from camera or static image file\n"
"Either the ELL model file, or the compiled model's Python module must be given,\n"
"using the --model or --compiled_model options respectively.\n"
"Example:\n"
" python demo.py categories.txt --compiled_model tutorial1/pi3/model1\n"
" python demo.py categories.txt --model model1.ell\n"
"This shows opencv window with image classified by the model using given labels"
)
# required arguments
arg_parser.add_argument(
"labels",
help=
"path to the labels file for evaluating the model, or comma separated list if using more than one model"
)
# options
arg_parser.add_argument("--save",
help="save images captured by the camera",
action='store_true')
arg_parser.add_argument(
"--threshold",
type=float,
help=
"threshold for the minimum prediction score. A lower threshold will show more prediction labels, but they have a higher chance of being completely wrong.",
default=helper.threshold)
arg_parser.add_argument(
"--bgr",
help=
"specify True if input data should be in BGR format (default False)",
default=helper.bgr)
arg_parser.add_argument(
"--nogui",
help="disable GUI to enable automated testing of a batch of images",
action='store_true')
arg_parser.add_argument(
"--iterations",
type=int,
help=
"when used with --nogui this tests multiple iterations of each image to get better timing information"
)
arg_parser.add_argument(
"--print_labels",
help="print predictions instead of drawing them on the image",
action='store_true')
# mutually exclusive options
group = arg_parser.add_mutually_exclusive_group()
group.add_argument("--camera",
type=int,
help="the camera id of the webcam",
default=0)
group.add_argument(
"--image",
help=
"path to an image file. If set, evaluates the model using the image, instead of a webcam"
)
group.add_argument(
"--folder",
help=
"path to an image folder. If set, evaluates the model using the images found there"
)
group2 = arg_parser.add_mutually_exclusive_group()
group2.add_argument("--model", help="path to a model file")
group2.add_argument("--compiled_model",
help="path to the compiled model's Python module")
args = arg_parser.parse_args()
if not args.compiled_model and not args.model:
print("### Error: Required one of --model or --compiled_model")
return
# setup some options on the demo helper
if args.save:
helper.save_images = True
helper.threshold = args.threshold
if args.iterations:
helper.iterations = args.iterations
helper.set_input(args.camera, args.folder, args.image)
if args.print_labels:
helper.print_labels = True
helper.bgr = args.bgr
helper.nogui = args.nogui
# initialize the labels and the model (either reference or compiled)
helper.load_model(args.labels, args.model, args.compiled_model)
lastPrediction = ""
help_prompt = 60
while (not helper.done()):
# Grab next frame
frame = helper.get_next_frame()
# Prepare the image to send to the model.
# This involves scaling to the required input dimension and re-ordering from BGR to RGB
data = helper.prepare_image_for_predictor(frame)
# Get the model to classify the image, by returning a list of probabilities for the classes it can detect
predictions = helper.predict(data)
# Get the (at most) top 5 predictions that meet our threshold. This is returned as a list of tuples,
# each with the text label and the prediction score.
top5 = helper.get_top_n_predictions(predictions, 5)
# Turn the top5 into a text string to display
text = ", ".join([
"(" + str(int(element[1] * 100)) + "%) " +
helper.get_label(element[0]) for element in top5
])
save = False
if text != lastPrediction:
print(text)
save = True
lastPrediction = text
# Draw the text on the frame
if helper.nogui or helper.print_labels:
if helper.new_frame:
if not text:
text = "unknown"
if helper.image_filename:
text = helper.image_filename + ": " + text
print(text)
else:
frameToShow = frame
helper.draw_label(frameToShow, text)
helper.draw_fps(frameToShow)
if help_prompt > 0:
help_prompt -= 1
helper.draw_footer(frameToShow,
"Press ESC to close this window")
# Show the new frame
helper.show_image(frameToShow, save)
helper.report_times()