def predict_image(network,
path,
only_classes=None,
ignore_classes=None,
save_path=None):
click.echo('Predicting {}...'.format(path), nl=False)
# Open and read the image to predict.
with tf.gfile.Open(path, 'rb') as f:
try:
image = Image.open(f).convert('RGB')
except (tf.errors.OutOfRangeError, OSError) as e:
click.echo()
click.echo('Error while processing {}: {}'.format(path, e))
return
# Run image through the network.
objects = network.predict_image(image)
# Filter the results according to the user input.
objects = filter_classes(objects,
only_classes=only_classes,
ignore_classes=ignore_classes)
# Save predicted image.
if save_path:
vis_objects(np.array(image), objects).save(save_path)
click.echo(' done.')
return objects
def predict(model_name):
if request.method == 'GET':
return jsonify(error='Use POST method to send image.'), 400
try:
image_array = get_image2()
except ValueError:
return jsonify(error='Missing image'), 400
except OSError:
return jsonify(error='Incompatible file type'), 400
total_predictions = request.args.get('total')
if total_predictions is not None:
try:
total_predictions = int(total_predictions)
except ValueError:
total_predictions = None
# Wait for the model to finish loading.
NETWORK_START_THREAD.join()
objects = PREDICTOR_NETWORK.predict_image(image_array)
image = request.files.get('image')
if not image:
raise ValueError
img = Image.open(image.stream).convert('RGB')
# vis_objects(np.array(image_array), objects).save("/tmp/luminoth/data.png")
vis_objects(np.array(img), objects).save("/tmp/luminoth/data.png")
global ouputObjects
ouputObjects = objects
objects = objects[:total_predictions]
return jsonify({'objects': objects})
def predict_video(network,
path,
only_classes=None,
ignore_classes=None,
save_path=None):
if save_path:
# We hardcode the video output to mp4 for the time being.
save_path = os.path.splitext(save_path)[0] + '.mp4'
try:
writer = skvideo.io.FFmpegWriter(save_path)
except AssertionError as e:
tf.logging.error(e)
tf.logging.error(
'Please install ffmpeg before making video predictions.')
exit()
else:
click.echo(
'Video not being saved. Note that for the time being, no JSON '
'output is being generated. Did you mean to specify `--save-path`?'
)
num_of_frames = int(skvideo.io.ffprobe(path)['video']['@nb_frames'])
video_progress_bar = click.progressbar(skvideo.io.vreader(path),
length=num_of_frames,
label='Predicting {}'.format(path))
colormap = build_colormap()
objects_per_frame = []
with video_progress_bar as bar:
try:
start_time = time.time()
for idx, frame in enumerate(bar):
# Run image through network.
objects = network.predict_image(frame)
# Filter the results according to the user input.
objects = filter_classes(objects,
only_classes=only_classes,
ignore_classes=ignore_classes)
objects_per_frame.append({'frame': idx, 'objects': objects})
# Draw the image and write it to the video file.
if save_path:
image = vis_objects(frame, objects, colormap=colormap)
writer.writeFrame(np.array(image))
stop_time = time.time()
click.echo('fps: {0:.1f}'.format(num_of_frames /
(stop_time - start_time)))
except RuntimeError as e:
click.echo() # Error prints next to progress bar otherwise.
click.echo('Error while processing {}: {}'.format(path, e))
if save_path:
click.echo('Partially processed video file saved in {}'.format(
save_path))
if save_path:
writer.close()
return objects_per_frame
def predict(model_name):
try:
# TODO ADD more models
if request.method == 'GET':
print(request.args)
#return jsonify(error='Use POST method to send image.', count=-1)
total_predictions = request.args.get('total', None)
min_prob = request.args.get('min_prob', None)
only_number = request.args.get('only_number', False)
id_task = request.args.get('id', False)
try:
image_array = get_image_url()
except ValueError:
return jsonify(error='Missing image', count=-2)
except OSError:
return jsonify(error='Incompatible file type', count=-3)
else: #POST
total_predictions = request.form.get('total', None)
min_prob = request.form.get('min_prob', None)
only_number = request.form.get('only_number', False)
id_task = request.form.get('id', False)
try:
image_array = get_image()
except ValueError:
return jsonify(error='Missing image', count=-2)
except OSError:
return jsonify(error='Incompatible file type', count=-3)
if total_predictions is not None:
try:
total_predictions = int(total_predictions)
except ValueError:
total_predictions = None
if min_prob is not None:
try:
min_prob = float(min_prob)
except ValueError:
min_prob = None
if only_number == "False":
only_number = None
if not id_task:
return jsonify(error='Missing task_id', count=-4)
# Wait for the model to finish loading.
NETWORK_START_THREAD.join()
objects = PREDICTOR_NETWORK.predict_image(image_array)
if min_prob:
objects = [obj for obj in objects if obj['prob'] >= min_prob]
if total_predictions:
objects = objects[:total_predictions]
# Save predicted image.
if SAVE_PATH_GLOBAL:
print("{}{}_Counted.jpg".format(SAVE_PATH_GLOBAL,id_task))
vis_objects(np.array(image_array), objects).save("{}{}_Counted.jpg".format(SAVE_PATH_GLOBAL,id_task))
if only_number:
return jsonify({'count': len(objects)})
return jsonify({'objects': objects, 'count': len(objects)})
except Exception as e:
return jsonify(error='Unkown error', data=str(e), count=-666)