Skip to content

Commit

Permalink
Merge pull request #76 from Gregor-Mendel-Institute/development
Browse files Browse the repository at this point in the history 
Development
  • Loading branch information
@MiroslavPolacek
MiroslavPolacek authored Sep 23, 2022
2 parents 1558049 + 1e59387 commit a4addef
Show file tree
Hide file tree
Showing 3 changed files with 106 additions and 86 deletions.
6 changes: 5 additions & 1 deletion .gitignore
View file
Original file line number Diff line number Diff line change
@@ -1,4 +1,5 @@
logs
data
datasets
RUNID
model_eval
Expand All @@ -7,14 +8,17 @@ laptop_docker_test.sh
laptop_singularity_test.sh
trg-imageprocessing_master.sif
image-processing_master.sif
cluster_time_images.sh

Dockerfile_with_Cracks
Docker_commands
TRG-process.def

*.pyc
*.ipynb
**.pyc
**.stdout
__pycache__
__pycache__/
postprocessingCracksRings_backup.py
postprocessingCracksRings_test.py
Train_onlyRing_test.sh
19 changes: 19 additions & 0 deletions CoreProcessingPipelineScripts/CNN/Mask_RCNN/postprocessing/cluster_time_images.sh
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,19 @@
#!/usr/bin/env bash

#SBATCH --nodes=1
#SBATCH --partition=g
#SBATCH --gres=gpu:RTX:1
#SBATCH --mem=32G
#SBATCH --qos=medium
#SBATCH --time=00-20:00:00
#SBATCH --output=Time_images.stdout

ml load anaconda3/2019.03
source activate ~/.conda/envs/TreeRingCNN

~/.conda/envs/TreeRingCNN/bin/python3 postprocessingCracksRings.py \
--dpi=13039 \
--run_ID=Image_time_test_debug \
--input=/groups/swarts/lab/DendroImages/CNN_test/AlexPOS/MEECNNPaperTreeringSupplementalInfo/Tiffs \
--weightRing=/users/miroslav.polacek/TRG_testing/TRG-ImageProcessing/CoreProcessingPipelineScripts/CNN/Mask_RCNN/logs/onlyring/mask_rcnn_treeringcrackscomb2_onlyring_0186.h5 \
--output_folder=/users/miroslav.polacek/Time_results \
167 changes: 82 additions & 85 deletions CoreProcessingPipelineScripts/CNN/Mask_RCNN/postprocessing/postprocessingCracksRings.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,19 +4,9 @@
Run detections for squares in a form of sliding window with certain overlap to prevent problems of having ring directly at the edge.
Run detection separately for the best model for Ring detection and the best for Crack.
Fuse all detections in one mask layer and consequently attach all of them to each other creating mask layer
of the size of original image. The detection confidance needs to be set in the config!
of the size of original image. The detection confidence needs to be set in the config!
Export JSON and POS files.
Print the image with mask over it.
FOR TESTING Mac
conda activate TreeRingCNNtest &&
cd /Users/miroslav.polacek/github/TRG-ImageProcessing/CoreProcessingPipelineScripts/CNN/Mask_RCNN/postprocessing &&
python3 postprocessingCracks.py --dpi=12926 --run_ID=RUN_ID_SOME_VALUE --input=/Users/miroslav.polacek/Pictures/whole_core_examples --weight=/Users/miroslav.polacek/github/TreeRingCracksCNN/Mask_RCNN/logs/treeringcrackscomb20201119T2220/mask_rcnn_treeringcrackscomb_0222.h5 --output_folder=/Users/miroslav.polacek/Documents/CNNTestRuns
FOR TESTING MANJARO
conda activate TreeRingCNN &&
cd /home/miroslavp/Github/TRG-ImageProcessing/CoreProcessingPipelineScripts/CNN/Mask_RCNN/postprocessing &&
python3 postprocessingCracksRings.py --dpi=12926 --run_ID=RUN_ID_SOME_VALUE --input=/home/miroslavp/Pictures/whole_core_examples --weightRing=/home/miroslavp/Github/TRG-ImageProcessing/CoreProcessingPipelineScripts/CNN/Mask_RCNN/logs/treeringcrackscomb2_onlyring20210121T1457/mask_rcnn_treeringcrackscomb2_onlyring_0186.h5 --weightCrack=/home/miroslavp/Github/TRG-ImageProcessing/CoreProcessingPipelineScripts/CNN/Mask_RCNN/logs/treeringcrackscomb2_onlycracks20210121T2224/mask_rcnn_treeringcrackscomb2_onlycracks_0522.h5 --output_folder=/home/miroslavp/Documents/CNNTestRuns
"""

#######################################################################
Expand All @@ -27,6 +17,7 @@
# Parse command line arguments
parser = argparse.ArgumentParser(
description='Segmentation of whole core')

## Compulsary arguments
parser.add_argument('--dpi', required=True,
help="DPI value for the image")
Expand All @@ -45,6 +36,7 @@
parser.add_argument('--output_folder', required=True,
metavar="/path/to/out/folder",
help="Path to output folder")

## Optional arguments
parser.add_argument('--weightCrack', required=False,
metavar="/path/to/weight/file",
Expand Down Expand Up @@ -76,36 +68,26 @@
#######################################################################
import os
import sys
import random
import math
import re
import cv2
import json
import time
import skimage
import pandas as pd
import numpy as np
#import tensorflow as tf
import matplotlib
import matplotlib.pyplot as plt
import matplotlib.patches as patches
import shapely
from shapely.geometry import box
from shapely.ops import nearest_points
import scipy
from scipy import optimize
from datetime import datetime
from operator import itemgetter

# Import Mask RCNN
ROOT_DIR = os.path.abspath("../")
print('ROOT_DIR', ROOT_DIR)
sys.path.append(ROOT_DIR) # To find local version of the library
from mrcnn import utils
from mrcnn import visualize
from mrcnn.src_get_centerline import get_centerline
import mrcnn.model as modellib
from mrcnn.model import log
from DetectionConfig import TreeRing_onlyRing
from DetectionConfig import TreeRing_onlyCracks

Expand Down Expand Up @@ -349,7 +331,7 @@ def clean_up_mask(mask, min_mask_overlap=3, is_ring=True):
#type(binary_mask)
uint8binary = binary_mask.astype(np.uint8).copy()

# Older version of openCV has slightly different syntax i adjusted for it here
# Older version of openCV has slightly different syntax I adjusted for it here
if int(cv2.__version__.split(".")[0]) < 4:
_, contours, _ = cv2.findContours(uint8binary,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE)
else:
Expand Down Expand Up @@ -566,6 +548,9 @@ def measure_contours(Multi_centerlines, image):
break
cutting_point = PlusMinus_index[i+1][1] + ((PlusMinus_index[i+2][1] - PlusMinus_index[i+1][1])/2)
cutting_point_detected = 1
#if cutting_point is immediately at the beggining of the sample ignore it
if cutting_point < imgheight*2: # if cutting point is within 2*image height it will be ignored
cutting_point_detected = 0

# Split sequence where it is crossing the middle
if cutting_point_detected==1:
Expand Down Expand Up @@ -944,79 +929,91 @@ def main():

for f in input_list:
if f.endswith('.tif') and f.replace('.tif', '') not in json_list:
print("Processing image: {}".format(f))
write_run_info("Processing image: {}".format(f))
image_path = os.path.join(input_path, f)
im_origin = skimage.io.imread(image_path)

# Define cropUpandDown, overlap and detection_rows values if they were not provided as arguments
if args.cropUpandDown is not None:
cropUpandDown = float(args.cropUpandDown)
else:
cropUpandDown = 0.17

if args.sliding_window_overlap is not None:
sliding_window_overlap = float(args.sliding_window_overlap)
else:
sliding_window_overlap = 0.75

if args.n_detection_rows is None or args.n_detection_rows==1:
detection_rows = 1
else:
detection_rows=int(args.n_detection_rows)

detected_mask = sliding_window_detection_multirow(image = im_origin,
detection_rows=detection_rows,
modelRing=modelRing,
modelCrack=modelCrack,
overlap = sliding_window_overlap,
cropUpandDown = cropUpandDown)

detected_mask_rings = detected_mask[:,:,0]
print("detected_mask_rings", detected_mask_rings.shape)
try:
image_start_time = time.time()
print("Processing image: {}".format(f))
write_run_info("Processing image: {}".format(f))
image_path = os.path.join(input_path, f)
im_origin = skimage.io.imread(image_path)

# Define default values if they were not provided as arguments
if args.cropUpandDown is not None:
cropUpandDown = float(args.cropUpandDown)
else:
cropUpandDown = 0.17

# Define minimum mask overlap if not provided
if args.min_mask_overlap is not None:
min_mask_overlap = args.min_mask_overlap
else:
min_mask_overlap = 3
if args.sliding_window_overlap is not None:
sliding_window_overlap = float(args.sliding_window_overlap)
else:
sliding_window_overlap = 0.75

clean_contours_rings = clean_up_mask(detected_mask_rings, min_mask_overlap=min_mask_overlap, is_ring=True)
if args.n_detection_rows is None or args.n_detection_rows==1:
detection_rows = 1
else:
detection_rows=int(args.n_detection_rows)

centerlines_rings = find_centerlines(clean_contours_rings)
detected_mask = sliding_window_detection_multirow(image = im_origin,
detection_rows=detection_rows,
modelRing=modelRing,
modelCrack=modelCrack,
overlap = sliding_window_overlap,
cropUpandDown = cropUpandDown)

centerlines, measure_points, cutting_point = measure_contours(centerlines_rings, detected_mask_rings)
detected_mask_rings = detected_mask[:,:,0]
print("detected_mask_rings", detected_mask_rings.shape)

# If cracks are detected
clean_contours_cracks = None
if args.weightCrack is not None:
detected_mask_cracks = detected_mask[:,:,1]
print("detected_mask_cracks", detected_mask_cracks.shape)
clean_contours_cracks = clean_up_mask(detected_mask_cracks, is_ring=False)
# Define minimum mask overlap if not provided
if args.min_mask_overlap is not None:
min_mask_overlap = int(args.min_mask_overlap)
else:
min_mask_overlap = 3

write_to_json(image_name=f,cutting_point=cutting_point, run_ID=run_ID,
path_out=path_out, centerlines_rings=centerlines_rings,
clean_contours_rings=clean_contours_rings,
clean_contours_cracks=clean_contours_cracks)
clean_contours_rings = clean_up_mask(detected_mask_rings, min_mask_overlap=min_mask_overlap, is_ring=True)

image_name = f.replace('.tif', '')
DPI = float(args.dpi)
write_to_pos(centerlines, measure_points, image_name, f, DPI, path_out)
centerlines_rings = find_centerlines(clean_contours_rings)

if args.print_detections == "yes":
# Ploting lines is moslty for debugging
write_run_info("Printing detection PNGs")
print("Printing detection PNGs")
masked_image = im_origin.astype(np.uint32).copy()
masked_image = apply_mask(masked_image, detected_mask_rings, alpha=0.2)
centerlines, measure_points, cutting_point = measure_contours(centerlines_rings, detected_mask_rings)

# If cracks are detected
clean_contours_cracks = None
if args.weightCrack is not None:
masked_image = apply_mask(masked_image, detected_mask_cracks, alpha=0.3)
detected_mask_cracks = detected_mask[:,:,1]
print("detected_mask_cracks", detected_mask_cracks.shape)
clean_contours_cracks = clean_up_mask(detected_mask_cracks, is_ring=False)

write_to_json(image_name=f,cutting_point=cutting_point, run_ID=run_ID,
path_out=path_out, centerlines_rings=centerlines_rings,
clean_contours_rings=clean_contours_rings,
clean_contours_cracks=clean_contours_cracks)

image_name = f.replace('.tif', '')
DPI = float(args.dpi)
write_to_pos(centerlines, measure_points, image_name, f, DPI, path_out)

if args.print_detections == "yes":
# Ploting lines is moslty for debugging
write_run_info("Printing detection PNGs")
print("Printing detection PNGs")
masked_image = im_origin.astype(np.uint32).copy()
masked_image = apply_mask(masked_image, detected_mask_rings, alpha=0.2)

if args.weightCrack is not None:
masked_image = apply_mask(masked_image, detected_mask_cracks, alpha=0.3)

plot_lines(masked_image, centerlines, measure_points,
image_name, path_out)
write_run_info("IMAGE FINISHED")
image_finished_time = time.time()
image_run_time = image_finished_time - image_start_time
write_run_info(f"Image run time: {image_run_time} s")
print("IMAGE FINISHED")

except Exception as e:
write_run_info(e)
write_run_info("IMAGE WAS NOT FINISHED")
print(e)
print("IMAGE WAS NOT FINISHED")

plot_lines(masked_image, centerlines, measure_points,
image_name, path_out)
write_run_info("IMAGE FINISHED")
print("IMAGE FINISHED")

if __name__ == '__main__':
main()

0 comments on commit a4addef

Please sign in to comment.