Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Sign up
Appearance settings

Commit 0eb619c

Browse files
Feat: add concurrency to findMatches methods
1 parent a3ed2ec commit 0eb619c

File tree

3 files changed

+88
-52
lines changed

3 files changed

+88
-52
lines changed

‎CHANGELOG.md

Lines changed: 5 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -3,6 +3,11 @@ All notable changes to this project will be documented in this file.
33

44
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
55

6+
### [1.6.4] - 2023年01月19日
7+
8+
### Changed
9+
- Improve speed by adding concurrency in the findMatches method, using half the number of cpu cores available.
10+
611
### [1.6.3] - 2021年11月24日
712

813
### Changed

‎MTM/__init__.py

Lines changed: 82 additions & 51 deletions
Original file line numberDiff line numberDiff line change
@@ -1,13 +1,16 @@
11
"""Main code for Multi-Template-Matching (MTM)."""
2+
import os
3+
import warnings
4+
from concurrent.futures import ThreadPoolExecutor, as_completed
5+
26
import cv2
3-
import numpy as np
7+
import numpy as np
48
import pandas as pd
5-
import warnings
9+
fromscipy.signalimport find_peaks
610
from skimage.feature import peak_local_max
7-
from scipy.signal import find_peaks
8-
from .version import __version__
911

1012
from .NMS import NMS
13+
from .version import __version__
1114

1215
__all__ = ['NMS']
1316

@@ -33,7 +36,7 @@ def _findLocalMax_(corrMap, score_threshold=0.6):
3336
peaks = [[i,0] for i in peaks[0]]
3437

3538

36-
else: # Correlatin map is 2D
39+
else: # Correlation map is 2D
3740
peaks = peak_local_max(corrMap, threshold_abs=score_threshold, exclude_border=False).tolist()
3841

3942
return peaks
@@ -116,82 +119,110 @@ def findMatches(listTemplates, image, method=cv2.TM_CCOEFF_NORMED, N_object=floa
116119
-------
117120
- Pandas DataFrame with 1 row per hit and column "TemplateName"(string), "BBox":(X, Y, Width, Height), "Score":float
118121
"""
119-
if N_object != float("inf") and type(N_object) !=int:
122+
if N_object != float("inf") and notisinstance(N_object, int):
120123
raise TypeError("N_object must be an integer")
121124

122125
## Crop image to search region if provided
123126
if searchBox is not None:
124127
xOffset, yOffset, searchWidth, searchHeight = searchBox
125128
image = image[yOffset : yOffset+searchHeight, xOffset : xOffset+searchWidth]
126-
127129
else:
128130
xOffset=yOffset=0
129-
131+
130132
# Check that the template are all smaller are equal to the image (original, or cropped if there is a search region)
131133
for index, tempTuple in enumerate(listTemplates):
132-
134+
133135
if not isinstance(tempTuple, tuple) or len(tempTuple)==1:
134136
raise ValueError("listTemplates should be a list of tuples as ('name','array') or ('name', 'array', 'mask')")
135-
137+
136138
templateSmallerThanImage = all(templateDim <= imageDim for templateDim, imageDim in zip(tempTuple[1].shape, image.shape))
137-
139+
138140
if not templateSmallerThanImage :
139141
fitIn = "searchBox" if (searchBox is not None) else "image"
140142
raise ValueError("Template '{}' at index {} in the list of templates is larger than {}.".format(tempTuple[0], index, fitIn) )
141-
143+
142144
listHit = []
143-
for tempTuple in listTemplates:
145+
## Use multi-threading to iterate through all templates, using half the number of cpu cores available.
146+
with ThreadPoolExecutor(max_workers=round(os.cpu_count()*.5)) as executor:
147+
futures = [executor.submit(_multi_compute, tempTuple, image, method, N_object, score_threshold, xOffset, yOffset, listHit) for tempTuple in listTemplates]
148+
for future in as_completed(futures):
149+
_ = future.result()
144150

145-
templateName, template = tempTuple[:2]
146-
mask = None
151+
if listHit:
152+
return pd.DataFrame(listHit) # All possible hits before Non-Maxima Supression
153+
else:
154+
return pd.DataFrame(columns=["TemplateName", "BBox", "Score"])
147155

148-
if len(tempTuple)>=3: # ie a mask is also provided
149-
if method in (0,3):
150-
mask = tempTuple[2]
151-
else:
152-
warnings.warn("Template matching method not supporting the use of Mask. Use 0/TM_SQDIFF or 3/TM_CCORR_NORMED.")
153156

154-
#print('\nSearch with template : ',templateName)
155-
corrMap = computeScoreMap(template, image, method, mask=mask)
157+
def _multi_compute(tempTuple, image, method, N_object, score_threshold, xOffset, yOffset, listHit):
158+
"""
159+
Find all possible template locations satisfying the score threshold provided a template to search and an image.
160+
Add the hits in the list of hits.
161+
162+
Parameters
163+
----------
164+
- tempTuple : a tuple (LabelString, template, mask (optional))
165+
template to search in each image, associated to a label
166+
labelstring : string
167+
template : numpy array (grayscale or RGB)
168+
mask (optional): numpy array, should have the same dimensions and type than the template
156169
157-
## Find possible location of the object
158-
if N_object==1: # Detect global Min/Max
159-
minVal, maxVal, minLoc, maxLoc = cv2.minMaxLoc(corrMap)
170+
- image : Grayscale or RGB numpy array
171+
image in which to perform the search, it should be the same bitDepth and number of channels than the templates
160172
161-
ifmethod in (0,1):
162-
peaks= [minLoc[::-1]] # opposite sorting than in the multiple detection
173+
- method : int
174+
one of OpenCV template matching method (0 to 5), default 5=0-mean cross-correlation
163175
164-
else:
165-
peaks= [maxLoc[::-1]]
176+
- N_object: int or float("inf")
177+
expected number of objects in the image, default to infinity if unknown
166178
179+
- score_threshold: float in range [0,1]
180+
if N_object>1, returns local minima/maxima respectively below/above the score_threshold
167181
168-
else:# Detect local max or min
169-
if method in (0,1): # Difference => look for local minima
170-
peaks = _findLocalMin_(corrMap, score_threshold)
182+
- xOffset : int
183+
optional the x offset if the search area is provided
171184
172-
else:
173-
peaks=_findLocalMax_(corrMap, score_threshold)
185+
- yOffset : int
186+
optional the y offset if the search area is provided
174187
188+
- listHit : the list of hits which we want to add the discovered hit
189+
expected array of hits
190+
"""
191+
templateName, template = tempTuple[:2]
192+
mask = None
175193

176-
#print('Initially found',len(peaks),'hit with this template')
194+
if len(tempTuple)>=3: # ie a mask is also provided
195+
if method in (0,3):
196+
mask = tempTuple[2]
197+
else:
198+
warnings.warn("Template matching method not supporting the use of Mask. Use 0/TM_SQDIFF or 3/TM_CCORR_NORMED.")
177199

200+
#print('\nSearch with template : ',templateName)
201+
corrMap = computeScoreMap(template, image, method, mask=mask)
178202

179-
# Once every peak was detected for this given template
180-
## Create a dictionnary for each hit with {'TemplateName':, 'BBox': (x,y,Width, Height), 'Score':coeff}
203+
## Find possible location of the object
204+
if N_object==1: # Detect global Min/Max
205+
_, _, minLoc, maxLoc = cv2.minMaxLoc(corrMap)
206+
if method in (0,1):
207+
peaks = [minLoc[::-1]] # opposite sorting than in the multiple detection
208+
else:
209+
peaks = [maxLoc[::-1]]
210+
else:# Detect local max or min
211+
if method in (0,1): # Difference => look for local minima
212+
peaks = _findLocalMin_(corrMap, score_threshold)
213+
else:
214+
peaks = _findLocalMax_(corrMap, score_threshold)
181215

182-
height, width= template.shape[0:2] # slicing make sure it works for RGB too
216+
#print('Initially found',len(peaks),'hit with this template')
183217

184-
for peak in peaks :
185-
coeff = corrMap[tuple(peak)]
186-
newHit = {'TemplateName':templateName, 'BBox': ( int(peak[1])+xOffset, int(peak[0])+yOffset, width, height ) , 'Score':coeff}
218+
# Once every peak was detected for this given template
219+
## Create a dictionnary for each hit with {'TemplateName':, 'BBox': (x,y,Width, Height), 'Score':coeff}
187220

188-
# append to list of potential hit before Non maxima suppression
189-
listHit.append(newHit)
221+
height, width = template.shape[0:2] # slicing make sure it works for RGB too
190222

191-
if listHit:
192-
return pd.DataFrame(listHit) # All possible hits before Non-Maxima Supression
193-
else:
194-
return pd.DataFrame(columns=["TemplateName", "BBox", "Score"]) # empty df with correct column header
223+
for peak in peaks :
224+
# append to list of potential hit before Non maxima suppression
225+
listHit.append({'TemplateName':templateName, 'BBox': ( int(peak[1])+xOffset, int(peak[0])+yOffset, width, height ) , 'Score':corrMap[tuple(peak)]}) # empty df with correct column header
195226

196227

197228
def matchTemplates(listTemplates, image, method=cv2.TM_CCOEFF_NORMED, N_object=float("inf"), score_threshold=0.5, maxOverlap=0.25, searchBox=None):
@@ -239,7 +270,7 @@ def matchTemplates(listTemplates, image, method=cv2.TM_CCOEFF_NORMED, N_object=f
239270
tableHit = findMatches(listTemplates, image, method, N_object, score_threshold, searchBox)
240271

241272
if method == 0: raise ValueError("The method TM_SQDIFF is not supported. Use TM_SQDIFF_NORMED instead.")
242-
sortAscending = Trueifmethod==1elseFalse
273+
sortAscending = (method==1)
243274

244275
return NMS(tableHit, score_threshold, sortAscending, N_object, maxOverlap)
245276

@@ -275,7 +306,7 @@ def drawBoxesOnRGB(image, tableHit, boxThickness=2, boxColor=(255, 255, 00), sho
275306
if image.ndim == 2: outImage = cv2.cvtColor(image, cv2.COLOR_GRAY2RGB) # convert to RGB to be able to show detections as color box on grayscale image
276307
else: outImage = image.copy()
277308

278-
for index, row in tableHit.iterrows():
309+
for _, row in tableHit.iterrows():
279310
x,y,w,h = row['BBox']
280311
cv2.rectangle(outImage, (x, y), (x+w, y+h), color=boxColor, thickness=boxThickness)
281312
if showLabel: cv2.putText(outImage, text=row['TemplateName'], org=(x, y), fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=labelScale, color=labelColor, lineType=cv2.LINE_AA)
@@ -315,9 +346,9 @@ def drawBoxesOnGray(image, tableHit, boxThickness=2, boxColor=255, showLabel=Fal
315346
if image.ndim == 3: outImage = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY) # convert to RGB to be able to show detections as color box on grayscale image
316347
else: outImage = image.copy()
317348

318-
for index, row in tableHit.iterrows():
349+
for _, row in tableHit.iterrows():
319350
x,y,w,h = row['BBox']
320351
cv2.rectangle(outImage, (x, y), (x+w, y+h), color=boxColor, thickness=boxThickness)
321352
if showLabel: cv2.putText(outImage, text=row['TemplateName'], org=(x, y), fontFace=cv2.FONT_HERSHEY_SIMPLEX, fontScale=labelScale, color=labelColor, lineType=cv2.LINE_AA)
322353

323-
return outImage
354+
return outImage

‎MTM/version.py

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -2,4 +2,4 @@
22
# 1) we don't load dependencies by storing it in __init__.py
33
# 2) we can import it in setup.py for the same reason
44
# 3) we can import it into your module module
5-
__version__ = '1.6.3'
5+
__version__ = '1.6.4'

0 commit comments

Comments
(0)

AltStyle によって変換されたページ (->オリジナル) /