Skip to content

Commit

Permalink
version changed, black
Browse files Browse the repository at this point in the history
  • Loading branch information
@RBhupi
RBhupi committed Nov 9, 2024
1 parent 66299b8 commit 3226959
Show file tree
Hide file tree
Showing 2 changed files with 153 additions and 115 deletions.
266 changes: 152 additions & 114 deletions app.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -37,29 +37,30 @@
import cv2



def upload_image(sky_curr, timestamp, thres_otsu, plugin):
"""
Upload an image to beehive.
This function uploads the sky_curr image, as a JPEG file.
If an exception occurs during file removal, the error message is published.
"""
img2_file_name = 'img2_' + str(timestamp) + '.jpg'
img2_file_name = "img2_" + str(timestamp) + ".jpg"
cv2.imwrite(img2_file_name, sky_curr)
plugin.upload_file(img2_file_name, meta={'thres_otsu': str(thres_otsu)}, timestamp=timestamp)
plugin.upload_file(
img2_file_name, meta={"thres_otsu": str(thres_otsu)}, timestamp=timestamp
)

try:
os.remove(img2_file_name)
except Exception as e:
plugin.publish('exit.error', e)
plugin.publish("exit.error", e)
sys.exit(-1)


def main(args):
""" Runs the entire CMV workflow.
Takes in input args and run the whole CMV workflow i.e. capture frames,
"""Runs the entire CMV workflow.
Takes in input args and run the whole CMV workflow i.e. capture frames,
compute optical flow, perform clustering, and publish data to beehive.
Args:
Expand All @@ -68,199 +69,236 @@ def main(args):
Returns:
None
"""


#Create a dictionary to save settings
# Create a dictionary to save settings
inf = getInfoDict(args)

with Plugin() as plugin:
#get video frame and crop info into the dictionary.
# get video frame and crop info into the dictionary.
with Camera(args.input) as camera:
inf = cropMarginInfo(camera, inf)
#Counting frames and time-steps for netcdf output requirment.

# Counting frames and time-steps for netcdf output requirment.
fcount = 0

with Camera(args.input) as camera:
try:
sample = camera.snapshot()
except Exception as e:
logging.error(f"Run-time error in First camera.snapshot: {e}")
plugin.publish('exit.error', e)
plugin.publish("exit.error", e)
sys.exit(-1)
frame_time_curr = sample.timestamp/10**9
frame_time_curr = sample.timestamp / 10**9
fcount, sky_curr = cropFrame(sample, fcount, inf)

run_on = True

while run_on:
if inf['interval'] > 0:
time.sleep(inf['interval'])
#read new frame
if inf["interval"] > 0:
time.sleep(inf["interval"])
# read new frame
with Camera(args.input) as camera:
try:
sample = camera.snapshot()
except Exception as e:
logging.error(f"Run-time error in Second camera.snapshot: {e}")
plugin.publish('exit.error', e)
plugin.publish("exit.error", e)
sys.exit(-1)
frame_time_new = sample.timestamp/10**9
frame_time_new = sample.timestamp / 10**9
fcount, sky_new = cropFrame(sample, fcount, inf)


frame_time_prev = frame_time_curr
frame_time_curr = frame_time_new

vel_factor = 60/(frame_time_curr-frame_time_prev)
vel_factor = 60 / (frame_time_curr - frame_time_prev)

sky_prev = sky_curr
sky_curr = sky_new

#comput optical flow
# with plugin.timeit("plg.inf.time_ns"):
flow = cv2.calcOpticalFlowFarneback(sky_prev, sky_curr, None,
pyr_scale=0.5, levels=3,
winsize=inf['winsize'],
iterations=3, poly_n=inf['poly_n'],
poly_sigma=inf['poly_s'], flags=0)



mag, ang = cv2.cartToPolar(flow[..., 0], flow[..., 1], angleInDegrees = False)
mag=mag * vel_factor
#Use threshold for small values to remove background noise

# comput optical flow
# with plugin.timeit("plg.inf.time_ns"):
flow = cv2.calcOpticalFlowFarneback(
sky_prev,
sky_curr,
None,
pyr_scale=0.5,
levels=3,
winsize=inf["winsize"],
iterations=3,
poly_n=inf["poly_n"],
poly_sigma=inf["poly_s"],
flags=0,
)

mag, ang = cv2.cartToPolar(flow[..., 0], flow[..., 1], angleInDegrees=False)
mag = mag * vel_factor
# Use threshold for small values to remove background noise
thres_otsu = np.round(threshold_otsu(mag))

plugin.publish('cmv.thresh.otsu', float(thres_otsu), timestamp=sample.timestamp)
plugin.publish(
"cmv.thresh.otsu", float(thres_otsu), timestamp=sample.timestamp
)

# If it crossed the max threshold, upload sample image
if thres_otsu > args.thr:
upload_image(sky_curr, sample.timestamp, thres_otsu, plugin)

thres_mag = thres_otsu
#Reset the threshold for better range
# Reset the threshold for better range
if thres_otsu < 2:
thres_mag = 2
if thres_otsu >10:
if thres_otsu > 10:
thres_mag = 10



mag_mask = np.repeat(mag[:, :, np.newaxis], 2, axis=2)

flow= np.ma.masked_where(mag_mask<thres_mag, flow)
flow = np.ma.masked_where(mag_mask < thres_mag, flow)

if np.ma.MaskedArray.count(flow)==0:
plugin.publish('cmv.motion.detected', int(-1), timestamp=sample.timestamp)
if np.ma.MaskedArray.count(flow) == 0:
plugin.publish(
"cmv.motion.detected", int(-1), timestamp=sample.timestamp
)
continue


#recompute the mag and angle
mag, ang = cv2.cartToPolar(flow[..., 0], flow[..., 1], angleInDegrees = True)
# recompute the mag and angle
mag, ang = cv2.cartToPolar(flow[..., 0], flow[..., 1], angleInDegrees=True)
mag = mag * vel_factor
ang = (90+ang) % 360

ang = (90 + ang) % 360

# Number of Superpixels
num_seg = args.segments

#Use HSV coding for clustering
# Use HSV coding for clustering
hsv = np.zeros([mag.shape[0], mag.shape[1], 3], dtype=np.float32)
hsv[..., 1] = 250 #255-curr_frame #250
hsv[..., 1] = 250 # 255-curr_frame #250

# Use Hue and Value to encode the Optical Flow
hsv[..., 0] = ang/2
hsv[..., 0] = ang / 2
hsv[..., 2] = cv2.normalize(mag, None, 0, 255, cv2.NORM_MINMAX)
# Convert HSV image to BGR
image = cv2.cvtColor(hsv, cv2.COLOR_HSV2BGR)

# High compactness = squared regions & higher sigma = rounded segments
# Achanta, R., Shaji, A., Smith, K., Lucchi, A., Fua, P., & Süsstrunk, S.
# Achanta, R., Shaji, A., Smith, K., Lucchi, A., Fua, P., & Süsstrunk, S.
# (2012). SLIC superpixels compared to state-of-the-art superpixel methods.
# IEEE transactions on pattern analysis and machine intelligence,
# IEEE transactions on pattern analysis and machine intelligence,
# 34(11), 2274-2282.
segments = slic(image, n_segments=num_seg, sigma=2, compactness=10, convert2lab=True)
segments = slic(
image, n_segments=num_seg, sigma=2, compactness=10, convert2lab=True
)
segments_found = segments.max()
# Superpixel average color
superpixels = color.label2rgb(segments, image, kind='avg')
superpixels = color.label2rgb(segments, image, kind="avg")

#first remove the zero vector region from segmentation
# first remove the zero vector region from segmentation
seg_count = np.bincount(segments.ravel())

motion_detected = 0

for i in range(0, seg_count.shape[0]):
seg_id = seg_count.argmax()
seg_size = seg_count.max()
mag_mean = np.mean(mag[segments==seg_id])
ang_mean = np.mean(ang[segments==seg_id])
mag_median = np.median(mag[segments==seg_id])
ang_median = np.median(ang[segments==seg_id])
mag_mean = np.mean(mag[segments == seg_id])
ang_mean = np.mean(ang[segments == seg_id])
mag_median = np.median(mag[segments == seg_id])
ang_median = np.median(ang[segments == seg_id])

# make it zero for next iteration
seg_count[seg_id] = 0
meta={'seg_id':str(seg_id),
'seg_size': str(seg_size),
'nsegments_found':str(segments_found),
'input': args.input,
'channel': str(args.c),
'image_frac': str(args.k),
'quality': str(args.q),
'nsegments_asked':str(args.segments)
}


#Publish the output
meta = {
"seg_id": str(seg_id),
"seg_size": str(seg_size),
"nsegments_found": str(segments_found),
"input": args.input,
"channel": str(args.c),
"image_frac": str(args.k),
"quality": str(args.q),
"nsegments_asked": str(args.segments),
}

# Publish the output
if not np.isnan(mag_mean) and float(mag_median) > thres_mag:
motion_detected = motion_detected + 1
meta['seg_rank']= str(motion_detected) # At this time `motions_detected` counter shows rank of the segment
meta["seg_rank"] = str(
motion_detected
) # At this time `motions_detected` counter shows rank of the segment

plugin.publish(
"cmv.mean.mag.pxpm",
np.round(mag_mean),
meta=meta,
timestamp=sample.timestamp,
)
plugin.publish(
"cmv.mean.dir.degn",
np.round(ang_mean),
meta=meta,
timestamp=sample.timestamp,
)
plugin.publish(
"cmv.median.mag.pxpm",
np.round(mag_median),
meta=meta,
timestamp=sample.timestamp,
)
plugin.publish(
"cmv.median.dir.degn",
np.round(ang_median),
meta=meta,
timestamp=sample.timestamp,
)
# print('thres={} \t mag={} angle={}, seg_size={}, seg_id={}'.format(thres_mag, float(mag_mean), int(ang_mean), seg_size, seg_id))

plugin.publish(
"cmv.motion.detected", motion_detected, timestamp=sample.timestamp
)

plugin.publish('cmv.mean.mag.pxpm', np.round(mag_mean), meta=meta, timestamp=sample.timestamp)
plugin.publish('cmv.mean.dir.degn', np.round(ang_mean), meta=meta, timestamp=sample.timestamp)
plugin.publish('cmv.median.mag.pxpm', np.round(mag_median), meta=meta, timestamp=sample.timestamp)
plugin.publish('cmv.median.dir.degn', np.round(ang_median), meta=meta, timestamp=sample.timestamp)
#print('thres={} \t mag={} angle={}, seg_size={}, seg_id={}'.format(thres_mag, float(mag_mean), int(ang_mean), seg_size, seg_id))

plugin.publish('cmv.motion.detected', motion_detected, timestamp=sample.timestamp)


if args.oneshot:
run_on = False





if __name__ == "__main__":
parser = argparse.ArgumentParser(description='''
parser = argparse.ArgumentParser(
description="""
This program uses optical flow method
from CV2 to compute the cloud
motion vectors in the hemispheric camera''')
parser.add_argument('--input', type=str,
help='Path to an input video or images.',
default="file://test-data/sgp-sage.mp4")
# default="/app/test-data/sgptsimovieS01.a1.20160726.000000.mpg")
parser.add_argument('--i', type=int,
help='Time interval in seconds.', default=30)
parser.add_argument('--c', type=int,
help='channels, 0=R, 1=G, 2=B, 9=Gr', default=0)
parser.add_argument('--k', type=float,
help='Keep fraction of the image after cropping', default=0.9)
parser.add_argument('--q', type=int,
help='''Quality of the motion field.
motion vectors in the hemispheric camera"""
)
parser.add_argument(
"--input",
type=str,
help="Path to an input video or images.",
default="file://test-data/sgp-sage.mp4",
)
# default="/app/test-data/sgptsimovieS01.a1.20160726.000000.mpg")
parser.add_argument("--i", type=int, help="Time interval in seconds.", default=30)
parser.add_argument(
"--c", type=int, help="channels, 0=R, 1=G, 2=B, 9=Gr", default=0
)
parser.add_argument(
"--k", type=float, help="Keep fraction of the image after cropping", default=0.9
)
parser.add_argument(
"--q",
type=int,
help="""Quality of the motion field.
Sets averaging window, poly_n and poly_sigma.
1-turbulant: detailed motion field but noisy.
2-smooth: less noisy and faster computation,''',
default=2)
parser.add_argument('--thr', type=int,
help='''Uploads images when magnitude is above this threshold''',
default=50)
parser.add_argument('--segments', type=int,
help=''' Number of Segments. ''',
default=100)
parser.add_argument('--oneshot', action= 'store_true',
help='''Run once and exit.''') #This is not working as intended when default option is used.


2-smooth: less noisy and faster computation,""",
default=2,
)
parser.add_argument(
"--thr",
type=int,
help="""Uploads images when magnitude is above this threshold""",
default=50,
)
parser.add_argument(
"--segments", type=int, help=""" Number of Segments. """, default=100
)
parser.add_argument(
"--oneshot", action="store_true", help="""Run once and exit."""
) # This is not working as intended when default option is used.

args = parser.parse_args()
main(args)
2 changes: 1 addition & 1 deletion sage.yaml
View file
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
name: "cloud-motion"
namespace: "waggle"
description: "Cloud Motion Estimator (Optical Flow) for the Sky Camera. Uploads images occasionally."
version: "1.23.07.12"
version: "1.24.11.8"
authors: "Bhupendra Raut <[email protected]>"
collaborators: "Dario Dematties Reyes, Joseph Swantek, Neal Conrad, Nicola Ferrier, Pete Beckman, Raj Sankaran, Robert Jackson, Scott Collis, Sean Shahkarami, Seongha Park, Sergey Shemyakin, Wolfgang Gerlach, Yongho Kim"
funding: "NSF 1935984"
Expand Down

0 comments on commit 3226959

Please sign in to comment.