Update morphocut.py
This commit is contained in:
tpollina
GitHub
parent
9d72a1e8bb
commit
70bf0a9680
|
|
@ -22,95 +22,148 @@ from morphocut.stat import RunningMedian
|
||||||
from morphocut.str import Format
|
from morphocut.str import Format
|
||||||
from morphocut.stream import TQDM, Enumerate
|
from morphocut.stream import TQDM, Enumerate
|
||||||
|
|
||||||
# import_path = "/data-ssd/mschroeder/Datasets/Pyrocystis_noctiluca/RAW"
|
from skimage.feature import canny
|
||||||
import_path = "/home/pi/Desktop/PlanktonScope_acquisition/01_16_2020/afternoon/14_1"
|
from skimage.color import rgb2gray, label2rgb
|
||||||
export_path = "/home/pi/Desktop/PlanktonScope_acquisition/01_16_2020/14_1_export"
|
from skimage.morphology import disk
|
||||||
archive_fn = os.path.join(export_path, "14_1_morphocut_processed.zip")
|
from skimage.morphology import erosion, dilation, closing
|
||||||
|
from skimage.measure import label, regionprops
|
||||||
|
|
||||||
|
|
||||||
|
import_path = "/media/tpollina/rootfs/home/pi/Desktop/PlanktonScope_acquisition/01_17_2020/RAW"
|
||||||
|
export_path = "/media/tpollina/rootfs/home/pi/Desktop/PlanktonScope_acquisition/01_17_2020/"
|
||||||
|
CLEAN = os.path.join(export_path, "CLEAN")
|
||||||
|
os.makedirs(CLEAN, exist_ok=True)
|
||||||
|
|
||||||
|
|
||||||
|
OBJECTS = os.path.join(export_path, "OBJECTS")
|
||||||
|
os.makedirs(OBJECTS, exist_ok=True)
|
||||||
|
|
||||||
|
archive_fn = os.path.join(export_path, "ecotaxa_export.zip")
|
||||||
|
|
||||||
# Meta data that is added to every object
|
# Meta data that is added to every object
|
||||||
global_metadata = {
|
global_metadata = {
|
||||||
"acq_instrument": "Planktoscope",
|
"acq_instrument": "Planktoscope",
|
||||||
"process_datetime": datetime.datetime.now(),
|
"process_datetime": datetime.datetime.now(),
|
||||||
|
"sample_project": "PlanktonScope Villefranche",
|
||||||
|
"sample_ship": "Kayak de Fabien",
|
||||||
|
"sample_operator": "Thibaut Pollina",
|
||||||
|
"sample_id": "Flowcam_PlanktonScope_comparison",
|
||||||
|
"sample_sampling_gear": "net",
|
||||||
|
"sample_time":150000,
|
||||||
|
"sample_date":16112020,
|
||||||
|
"object_lat": 43.696146,
|
||||||
|
"object_lon": 7.308359,
|
||||||
|
"acq_fnumber_objective": 16,
|
||||||
|
"acq_celltype": 200,
|
||||||
|
"process_pixel": 1.19,
|
||||||
|
"acq_camera": "Pi Camera V2.1",
|
||||||
|
"acq_instrument": "PlanktonScope V2.1",
|
||||||
|
"acq_software": "Node-RED Dashboard and raw python",
|
||||||
|
"acq_instrument_ID": "copepode",
|
||||||
|
"acq_volume": 24,
|
||||||
|
"acq_flowrate": "Unknown",
|
||||||
|
"acq_camera.resolution" : "(3280, 2464)",
|
||||||
|
"acq_camera.iso" : 60,
|
||||||
|
"acq_camera.shutter_speed" : 100,
|
||||||
|
"acq_camera.exposure_mode" : 'off',
|
||||||
|
"acq_camera.awb_mode" : 'off',
|
||||||
|
"acq_nb_frames" : 1000
|
||||||
}
|
}
|
||||||
|
|
||||||
if __name__ == "__main__":
|
# Define processing pipeline
|
||||||
print("Processing images under {}...".format(import_path))
|
with Pipeline() as p:
|
||||||
|
# Recursively find .jpg files in import_path.
|
||||||
|
# Sort to get consective frames.
|
||||||
|
abs_path = Find(import_path, [".jpg"], sort=True, verbose=True)
|
||||||
|
|
||||||
# Create export_path in case it doesn't exist
|
# Extract name from abs_path
|
||||||
os.makedirs(export_path, exist_ok=True)
|
name = Call(lambda p: os.path.splitext(os.path.basename(p))[0], abs_path)
|
||||||
|
|
||||||
# Define processing pipeline
|
# Read image
|
||||||
with Pipeline() as p:
|
img = ImageReader(abs_path)
|
||||||
# Recursively find .jpg files in import_path.
|
|
||||||
# Sort to get consective frames.
|
|
||||||
abs_path = Find(import_path, [".jpg"], sort=True, verbose=True)
|
|
||||||
|
|
||||||
# Extract name from abs_path
|
# Apply running median to approximate the background image
|
||||||
name = Call(lambda p: os.path.splitext(os.path.basename(p))[0], abs_path)
|
flat_field = RunningMedian(img, 10)
|
||||||
|
|
||||||
# Read image
|
# Correct image
|
||||||
img = ImageReader(abs_path)
|
img = img / flat_field
|
||||||
|
|
||||||
# Apply running median to approximate the background image
|
# Rescale intensities and convert to uint8 to speed up calculations
|
||||||
flat_field = RunningMedian(img, 10)
|
img = RescaleIntensity(img, in_range=(0, 1.1), dtype="uint8")
|
||||||
|
|
||||||
# Correct image
|
# Convert image to uint8 gray
|
||||||
img = img / flat_field
|
img_gray = RGB2Gray(img)
|
||||||
|
|
||||||
# Rescale intensities and convert to uint8 to speed up calculations
|
img_gray = Call(img_as_ubyte, img_gray)
|
||||||
img = RescaleIntensity(img, in_range=(0, 1.1), dtype="uint8")
|
|
||||||
|
|
||||||
# Show progress bar for frames
|
img_canny = Call(canny, img_gray, sigma=0.3)
|
||||||
TQDM(Format("Frame {name}", name=name))
|
|
||||||
|
|
||||||
# Convert image to uint8 gray
|
img_dilate = Call(dilation, img_canny)
|
||||||
img_gray = RGB2Gray(img)
|
|
||||||
img_gray = Call(img_as_ubyte, img_gray)
|
|
||||||
|
|
||||||
# Apply threshold find objects
|
img_closing = Call(closing, img_dilate)
|
||||||
threshold = 204 # Call(skimage.filters.threshold_otsu, img_gray)
|
|
||||||
mask = img_gray < threshold
|
|
||||||
|
|
||||||
# Write corrected frames
|
mask = Call(erosion, img_closing)
|
||||||
frame_fn = Format(os.path.join(export_path, "{name}.jpg"), name=name)
|
|
||||||
ImageWriter(frame_fn, img)
|
|
||||||
|
|
||||||
# Find objects
|
# Show progress bar for frames
|
||||||
regionprops = FindRegions(
|
TQDM(Format("Frame {name}", name=name))
|
||||||
mask, img_gray, min_area=100, padding=10, warn_empty=name
|
|
||||||
)
|
|
||||||
|
|
||||||
# For an object, extract a vignette/ROI from the image
|
# Apply threshold find objects
|
||||||
roi_orig = ExtractROI(img, regionprops, bg_color=255)
|
#threshold = 204 # Call(skimage.filters.threshold_otsu, img_gray)
|
||||||
roi_gray = ExtractROI(img_gray, regionprops, bg_color=255)
|
#mask = img_gray < threshold
|
||||||
|
|
||||||
# Generate an object identifier
|
# Write corrected frames
|
||||||
i = Enumerate()
|
frame_fn = Format(os.path.join(CLEAN, "{name}.jpg"), name=name)
|
||||||
object_id = Format("{name}_{i:d}", name=name, i=i)
|
ImageWriter(frame_fn, img)
|
||||||
|
|
||||||
# Calculate features. The calculated features are added to the global_metadata.
|
# Find objects
|
||||||
# Returns a Variable representing a dict for every object in the stream.
|
regionprops = FindRegions(
|
||||||
meta = CalculateZooProcessFeatures(
|
mask, img_gray, min_area=1000, padding=10, warn_empty=name
|
||||||
regionprops, prefix="object_", meta=global_metadata
|
)
|
||||||
)
|
|
||||||
# If CalculateZooProcessFeatures is not used, we need to copy global_metadata into the stream:
|
|
||||||
# meta = Call(lambda: global_metadata.copy())
|
|
||||||
# https://github.com/morphocut/morphocut/issues/51
|
|
||||||
|
|
||||||
# Add object_id to the metadata dictionary
|
# For an object, extract a vignette/ROI from the image
|
||||||
meta["object_id"] = object_id
|
roi_orig = ExtractROI(img, regionprops, bg_color=255)
|
||||||
|
|
||||||
# Generate object filenames
|
roi_orig
|
||||||
orig_fn = Format("{object_id}.jpg", object_id=object_id)
|
# Generate an object identifier
|
||||||
gray_fn = Format("{object_id}-gray.jpg", object_id=object_id)
|
i = Enumerate()
|
||||||
|
#Call(print,i)
|
||||||
|
|
||||||
# Write objects to an EcoTaxa archive:
|
object_id = Format("{name}_{i:d}", name=name, i=i)
|
||||||
# roi image in original color, roi image in grayscale, metadata associated with each object
|
#Call(print,object_id)
|
||||||
EcotaxaWriter(archive_fn, [(orig_fn, roi_orig), (gray_fn, roi_gray)], meta)
|
|
||||||
|
|
||||||
# Progress bar for objects
|
object_fn = Format(os.path.join(OBJECTS, "{name}.jpg"), name=object_id)
|
||||||
TQDM(Format("Object {object_id}", object_id=object_id))
|
ImageWriter(object_fn, roi_orig)
|
||||||
|
|
||||||
# Execute pipeline
|
|
||||||
p.run()
|
|
||||||
|
|
||||||
|
# Calculate features. The calculated features are added to the global_metadata.
|
||||||
|
# Returns a Variable representing a dict for every object in the stream.
|
||||||
|
meta = CalculateZooProcessFeatures(
|
||||||
|
regionprops, prefix="object_", meta=global_metadata
|
||||||
|
)
|
||||||
|
# If CalculateZooProcessFeatures is not used, we need to copy global_metadata into the stream:
|
||||||
|
# meta = Call(lambda: global_metadata.copy())
|
||||||
|
# https://github.com/morphocut/morphocut/issues/51
|
||||||
|
|
||||||
|
# Add object_id to the metadata dictionary
|
||||||
|
meta["object_id"] = object_id
|
||||||
|
|
||||||
|
# Generate object filenames
|
||||||
|
orig_fn = Format("{object_id}.jpg", object_id=object_id)
|
||||||
|
|
||||||
|
# Write objects to an EcoTaxa archive:
|
||||||
|
# roi image in original color, roi image in grayscale, metadata associated with each object
|
||||||
|
EcotaxaWriter(archive_fn, (orig_fn, roi_orig), meta)
|
||||||
|
|
||||||
|
# Progress bar for objects
|
||||||
|
TQDM(Format("Object {object_id}", object_id=object_id))
|
||||||
|
|
||||||
|
|
||||||
|
import datetime
|
||||||
|
|
||||||
|
BEGIN = datetime.datetime.now()
|
||||||
|
# Execute pipeline
|
||||||
|
p.run()
|
||||||
|
|
||||||
|
END = datetime.datetime.now()
|
||||||
|
|
||||||
|
print("MORPHOCUT :"+str(END-BEGIN))
|
||||||
|
|
|
||||||
Loading…
Reference in a new issue