pyscript-bootstrap-templates/examples/04_cell_detector/main.py

214 lines
6.7 KiB
Python

from pyscript_bootstrap_templates import bootstrap_templates
from pyscript_bootstrap_templates import bootstrap_HTML as bHTML
from pyscript_bootstrap_templates import bootstrap_inputs as bInputs
from pyscript_bootstrap_templates import HTML as HTML
import numpy as np
from PIL import Image
import cv2
loaded_img = None
app = bootstrap_templates.PyScriptBootstrapDashboard(parent_element="pyscript_app", brand_name="Pyscript Cell Detector")
main_div = bHTML.BootstrapContainer(parent=app.main)
main_div.w = 100
result_div = bHTML.BootstrapContainer(parent=main_div)
def process_image(image,
hough_min_dist = 500,
hough_param1 = 80,
hough_param2 = 500,
minRadius = 100,
maxRadius=500,
inner_hough_circles=True):
for child in result_div.children:
child.destroy()
#image = cv2.imread(str(test_img_path), cv2.IMREAD_COLOR)
# Convert the image to grayscale
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# Use median blur to reduce noise
gray = cv2.medianBlur(gray, 5)
# Apply Hough transform on the image to find circles
circles = cv2.HoughCircles(gray,
cv2.HOUGH_GRADIENT,
2,
minDist=hough_min_dist,
param1=hough_param1,
param2=hough_param2,
minRadius=minRadius,
maxRadius=maxRadius)
if circles is None:
bHTML.AlertDanger("No cell areas detected in image", parent=app.main)
return False
# Convert to integers
circles = np.uint16(np.around(circles))
# Loop over all detected circles and add them to the image
#for i in circles[0,:]:
# # Draw outer circle
# cv2.circle(image,(i[0],i[1]),i[2],(0,255,0),2)
# # Draw center of circle
# cv2.circle(image,(i[0],i[1]),2,(0,0,255),3)
first_wells = []
first_wells_gray = []
for i in circles[0,:]:
center = (i[0], i[1])
radius = i[2]
# Cut out the well
well = image[center[1]-radius:center[1]+radius, center[0]-radius:center[0]+radius]
well_gray = gray[center[1]-radius:center[1]+radius, center[0]-radius:center[0]+radius]
first_wells.append(well)
first_wells_gray.append(well_gray)
if inner_hough_circles:
second_wells = []
second_wells_gray = []
for well, well_gray in zip(first_wells, first_wells_gray):
circles = cv2.HoughCircles(well_gray, cv2.HOUGH_GRADIENT, 1, minDist=hough_min_dist, param1=50, param2=50, minRadius=(well.shape[0] // 2) - 50, maxRadius=(well.shape[0] // 2) - 10)
if circles is not None:
circles = np.uint16(np.around(circles))
i = circles[0,:][0]
center = (i[0], i[1])
radius = i[2] - 1
min_y = max(center[1]-radius, 0)
max_y = center[1]+radius
min_x = max(center[0]-radius, 0)
max_x = center[0]+radius
second_wells.append(well[min_y:max_y, min_x:max_x])
second_wells_gray.append(well_gray[min_y:max_y, min_x:max_x])
else:
second_wells.append(well)
second_wells_gray.append(well_gray)
else:
second_wells = first_wells
second_wells_gray = first_wells_gray
for well, well_gray in zip(second_wells, second_wells_gray):
_, binary = cv2.threshold(255 - well_gray, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
# Perform some morphological operations to remove small noise - you can change the kernel size
kernel = np.ones((3,3),np.uint8)
cleaned = cv2.morphologyEx(binary, cv2.MORPH_OPEN, kernel)
radius = well.shape[0] // 2 -1
# Now clear all pixels outside the circular well.
# We do this by creating a mask for the well and applying it to the image.
well_mask = np.zeros_like(cleaned, dtype=np.uint8)
cv2.circle(well_mask, (radius, radius), radius, 1, thickness=-1)
cleaned = cleaned * well_mask
circle = well_mask > 0
# now create an image overlay to display
ratio = np.sum((cleaned > 0).astype(int)) / np.sum(circle.astype(int))
div_result = bHTML.BootstrapContainer(f"ratio: {ratio * 100}%", parent=result_div)
div_result.shadow = bHTML.Shadow.LARGE
div_result.rounded = True
div_result.w = 50
well[cleaned > 0,0] = 255
final_img = bHTML.Image.from_numpy_array(well, parent=div_result)
final_img.rounded = True
final_img.rounded_size = 10
final_img.shadow = bHTML.Shadow.LARGE
final_img.w = 100
div = bHTML.BootstrapContainer("Controls", parent=app.sidebar)
div.font_size = 4
image_input = bInputs.InputFile(label_text="choose image file", parent=app.sidebar)
i_hough_min_dist = bInputs.InputInt("hough min distance [px]", parent=app.sidebar)
i_hough_min_dist.value = 500
i_hough_param1 = bInputs.InputInt("hough param1", parent=app.sidebar)
i_hough_param1.value = 80
i_hough_param2 = bInputs.InputInt("hough param2", parent=app.sidebar)
i_hough_param2.value = 500
i_hough_min_radius = bInputs.InputInt("hough min radius [px]", parent=app.sidebar)
i_hough_min_radius.value = 100
i_hough_max_radius = bInputs.InputInt("hough max radius [px]", parent=app.sidebar)
i_hough_max_radius.value = 500
btn = bHTML.ButtonPrimary("Process", parent=app.sidebar)
btn.w = 100
def on_image_change(f, *args):
global loaded_img
f.seek(0)
img = np.array(Image.open(f))
for child in result_div.children:
child.destroy()
output = bHTML.Image.from_numpy_array(img, parent=result_div)
output.rounded = True
output.rounded_size = 10
output.shadow = bHTML.Shadow.LARGE
output.w=50
loaded_img = img
image_input.onchange = on_image_change
def on_click(*args, **kwargs):
if loaded_img is None:
bHTML.AlertDanger("No image loaded", parent=app.main)
return
h_min_dist = int(i_hough_min_dist.value)
h_param1 = int(i_hough_param1.value)
h_param2 = int(i_hough_param2.value)
h_min_radius = int(i_hough_min_radius.value)
h_max_radius = int(i_hough_max_radius.value)
process_image(loaded_img,
hough_min_dist=h_min_dist,
hough_param1=h_param1,
hough_param2=h_param2,
minRadius=h_min_radius,
maxRadius=h_max_radius,
inner_hough_circles=True)
btn.onclick = on_click