feat: ImagePHashDatabase

This commit is contained in:
283375 2023-09-27 18:04:38 +08:00
parent 9c06c6d9f1
commit 34e56395ab
Signed by: 283375
SSH Key Fingerprint: SHA256:UcX0qg6ZOSDOeieKPGokA5h7soykG61nz2uxuQgVLSk
6 changed files with 159 additions and 72 deletions

View File

@ -25,7 +25,9 @@ DEVICES_JSON_FILE = "Ocr/DevicesJsonFile"
DEVICE_UUID = "Ocr/DeviceUuid"
TESSERACT_FILE = "Ocr/TesseractFile"
KNN_MODEL_FILE = "Ocr/KnnModelFile"
B30_KNN_MODEL_FILE = "Ocr/B30KnnModelFile"
SIFT_DATABASE_FILE = "Ocr/SiftDatabaseFile"
PHASH_DATABASE_FILE = "Ocr/PHashDatabaseFile"
ANDREAL_FOLDER = "Andreal/AndrealFolder"
ANDREAL_EXECUTABLE = "Andreal/AndrealExecutable"
@ -104,6 +106,15 @@ class Settings(QSettings, metaclass=QObjectSingleton):
def resetKnnModelFile(self):
self._resetStrItem(KNN_MODEL_FILE)
def b30KnnModelFile(self):
return self._strItem(B30_KNN_MODEL_FILE)
def setB30KnnModelFile(self, value: str):
self._setStrItem(B30_KNN_MODEL_FILE, value)
def resetB30KnnModelFile(self):
self._resetStrItem(B30_KNN_MODEL_FILE)
def siftDatabaseFile(self):
return self._strItem(SIFT_DATABASE_FILE)
@ -113,6 +124,15 @@ class Settings(QSettings, metaclass=QObjectSingleton):
def resetSiftDatabaseFile(self):
self._resetStrItem(SIFT_DATABASE_FILE)
def phashDatabaseFile(self):
return self._strItem(PHASH_DATABASE_FILE)
def setPHashDatabaseFile(self, value: str):
self._setStrItem(PHASH_DATABASE_FILE, value)
def resetPHashDatabaseFile(self):
self._resetStrItem(PHASH_DATABASE_FILE)
def andrealFolder(self):
return self._strItem(ANDREAL_FOLDER)

View File

@ -45,6 +45,17 @@ class SettingsOcr(SettingsBaseWidget):
self.knnModelFileResetButton,
)
if self.settings.b30KnnModelFile():
self.b30KnnModelFileValueWidget.selectFile(self.settings.b30KnnModelFile())
self.b30KnnModelFileValueWidget.filesSelected.connect(self.setB30KnnModelFile)
self.b30KnnModelFileResetButton.clicked.connect(self.resetB30KnnModelFile)
self.insertItem(
"b30KnnModelFile",
self.b30KnnModelFileLabel,
self.b30KnnModelFileValueWidget,
self.b30KnnModelFileResetButton,
)
if self.settings.siftDatabaseFile():
self.siftDatabaseFileValueWidget.selectFile(
self.settings.siftDatabaseFile()
@ -58,6 +69,21 @@ class SettingsOcr(SettingsBaseWidget):
self.siftDatabaseFileResetButton,
)
if self.settings.phashDatabaseFile():
self.phashDatabaseFileValueWidget.selectFile(
self.settings.phashDatabaseFile()
)
self.phashDatabaseFileValueWidget.filesSelected.connect(
self.setPHashDatabaseFile
)
self.phashDatabaseFileResetButton.clicked.connect(self.resetPHashDatabaseFile)
self.insertItem(
"phashDatabaseFile",
self.phashDatabaseFileLabel,
self.phashDatabaseFileValueWidget,
self.phashDatabaseFileResetButton,
)
def setDevicesJson(self):
selectedFile = self.devicesJsonValueWidget.selectedFiles()
if selectedFile and selectedFile[0]:
@ -79,8 +105,7 @@ class SettingsOcr(SettingsBaseWidget):
self.settings.resetDevicesJsonFile()
def setDeviceUuid(self):
device = self.deviceUuidValueWidget.currentData()
if device:
if device := self.deviceUuidValueWidget.currentData():
self.settings.setDeviceUuid(device.uuid)
def resetDeviceUuid(self):
@ -97,6 +122,16 @@ class SettingsOcr(SettingsBaseWidget):
self.knnModelFileValueWidget.reset()
self.settings.resetKnnModelFile()
def setB30KnnModelFile(self):
selectedFile = self.b30KnnModelFileValueWidget.selectedFiles()
if selectedFile and selectedFile[0]:
file = selectedFile[0]
self.settings.setB30KnnModelFile(file)
def resetB30KnnModelFile(self):
self.b30KnnModelFileValueWidget.reset()
self.settings.resetB30KnnModelFile()
def setSiftDatabaseFile(self):
selectedFile = self.siftDatabaseFileValueWidget.selectedFiles()
if selectedFile and selectedFile[0]:
@ -107,6 +142,16 @@ class SettingsOcr(SettingsBaseWidget):
self.siftDatabaseFileValueWidget.reset()
self.settings.resetSiftDatabaseFile()
def setPHashDatabaseFile(self):
selectedFile = self.phashDatabaseFileValueWidget.selectedFiles()
if selectedFile and selectedFile[0]:
file = selectedFile[0]
self.settings.setPHashDatabaseFile(file)
def resetPHashDatabaseFile(self):
self.phashDatabaseFileValueWidget.reset()
self.settings.resetPHashDatabaseFile()
def setupUi(self, *args):
self.devicesJsonLabel = QLabel(self)
self.devicesJsonValueWidget = FileSelector(self)
@ -120,10 +165,18 @@ class SettingsOcr(SettingsBaseWidget):
self.knnModelFileValueWidget = FileSelector(self)
self.knnModelFileResetButton = QPushButton(self)
self.b30KnnModelFileLabel = QLabel(self)
self.b30KnnModelFileValueWidget = FileSelector(self)
self.b30KnnModelFileResetButton = QPushButton(self)
self.siftDatabaseFileLabel = QLabel(self)
self.siftDatabaseFileValueWidget = FileSelector(self)
self.siftDatabaseFileResetButton = QPushButton(self)
self.phashDatabaseFileLabel = QLabel(self)
self.phashDatabaseFileValueWidget = FileSelector(self)
self.phashDatabaseFileResetButton = QPushButton(self)
super().setupUi(self)
self.retranslateUi()
@ -142,6 +195,12 @@ class SettingsOcr(SettingsBaseWidget):
self.knnModelFileLabel.setText(QCoreApplication.translate("Settings", "ocr.knnModelFile.label"))
self.knnModelFileResetButton.setText(QCoreApplication.translate("Settings", "resetButton"))
self.b30KnnModelFileLabel.setText(QCoreApplication.translate("Settings", "ocr.b30KnnModelFile.label"))
self.b30KnnModelFileResetButton.setText(QCoreApplication.translate("Settings", "resetButton"))
self.siftDatabaseFileLabel.setText(QCoreApplication.translate("Settings", "ocr.siftDatabaseFile.label"))
self.siftDatabaseFileResetButton.setText(QCoreApplication.translate("Settings", "resetButton"))
self.phashDatabaseFileLabel.setText(QCoreApplication.translate("Settings", "ocr.phashDatabaseFile.label"))
self.phashDatabaseFileResetButton.setText(QCoreApplication.translate("Settings", "resetButton"))
# fmt: on

View File

@ -12,7 +12,11 @@ from ui.designer.tabs.tabOcr.tabOcr_B30_ui import Ui_TabOcr_B30
from ui.extends.components.ocrQueue import OcrQueueModel
from ui.extends.shared.cv2_utils import cv2BgrMatToQImage, qImageToCvMatBgr
from ui.extends.shared.language import LanguageChangeEventFilter
from ui.extends.shared.settings import KNN_MODEL_FILE, SIFT_DATABASE_FILE, Settings
from ui.extends.shared.settings import (
B30_KNN_MODEL_FILE,
KNN_MODEL_FILE,
PHASH_DATABASE_FILE,
)
from ui.extends.tabs.tabOcr.tabOcr_B30 import ChieriV4OcrRunnable, b30ResultToScore
logger = logging.getLogger(__name__)
@ -35,12 +39,8 @@ class TabOcr_B30(Ui_TabOcr_B30, QWidget):
self.imageSelector.filesSelected.connect(self.imageSelected)
self.knnModelSelector.filesSelected.connect(self.knnModelSelected)
self.b30KnnModelSelector.filesSelected.connect(self.b30KnnModelSelected)
# self.siftDatabaseSelector.filesSelected.connect(self.siftDatabaseSelected)
self.phashDatabaseSelector.filesSelected.connect(self.phashDatabaseSelected)
self.knnModelSelector.connectSettings(KNN_MODEL_FILE)
# self.siftDatabaseSelector.connectSettings(SIFT_DATABASE_FILE)
self.imagePath = None # for checking only
self.img = None
self.paddleFolder = None
@ -54,10 +54,10 @@ class TabOcr_B30(Ui_TabOcr_B30, QWidget):
self.tryPrepareOcr.connect(self.prepareOcr)
settings = Settings()
logger.info("Applying default settings...")
self.knnModelSelector.selectFile(settings.knnModelFile())
# self.siftDatabaseSelector.selectFile(settings.siftDatabaseFile())
self.knnModelSelector.connectSettings(KNN_MODEL_FILE)
self.b30KnnModelSelector.connectSettings(B30_KNN_MODEL_FILE)
self.phashDatabaseSelector.connectSettings(PHASH_DATABASE_FILE)
self.ocrQueueModel = OcrQueueModel(self)
self.ocrQueue.setModel(self.ocrQueueModel)
@ -81,12 +81,6 @@ class TabOcr_B30(Ui_TabOcr_B30, QWidget):
self.b30KnnModel = cv2.ml.KNearest.load(b30KnnModelPath)
self.tryPrepareOcr.emit()
def siftDatabaseSelected(self):
if selectedFiles := self.siftDatabaseSelector.selectedFiles():
siftDatabasePath = selectedFiles[0]
self.siftDatabase = SIFTDatabase(siftDatabasePath)
self.tryPrepareOcr.emit()
def phashDatabaseSelected(self):
if selectedFiles := self.phashDatabaseSelector.selectedFiles():
phashDatabasePath = selectedFiles[0]

View File

@ -5,8 +5,8 @@ import cv2
# from arcaea_offline_ocr_device_creation_wizard.implements.wizard import Wizard
from arcaea_offline_ocr.device.v1.definition import DeviceV1
from arcaea_offline_ocr.device.v2.definition import DeviceV2
from arcaea_offline_ocr.sift_db import SIFTDatabase
from arcaea_offline_ocr.phash_db import ImagePHashDatabase
from arcaea_offline_ocr.sift_db import SIFTDatabase
from PySide6.QtCore import Qt, Slot
from PySide6.QtWidgets import QApplication, QFileDialog, QWidget
@ -16,7 +16,8 @@ from ui.extends.shared.language import LanguageChangeEventFilter
from ui.extends.shared.settings import (
DEVICES_JSON_FILE,
KNN_MODEL_FILE,
SIFT_DATABASE_FILE,
PHASH_DATABASE_FILE,
TESSERACT_FILE,
Settings,
)
from ui.extends.tabs.tabOcr.tabOcr_Device import (
@ -39,20 +40,15 @@ class TabOcr_Device(Ui_TabOcr_Device, QWidget):
self.deviceFileSelector.filesSelected.connect(self.deviceFileSelected)
self.knnModelSelector.filesSelected.connect(self.knnModelFileSelected)
# self.siftDatabaseSelector.filesSelected.connect(self.siftDatabaseFileSelected)
self.phashDatabaseSelector.filesSelected.connect(self.phashDatabaseFileSelected)
logger.info("Applying settings...")
self.deviceFileSelector.connectSettings(DEVICES_JSON_FILE)
self.knnModelSelector.connectSettings(KNN_MODEL_FILE)
# self.siftDatabaseSelector.connectSettings(SIFT_DATABASE_FILE)
self.tesseractFileSelector.connectSettings(TESSERACT_FILE)
self.phashDatabaseSelector.connectSettings(PHASH_DATABASE_FILE)
settings = Settings()
logger.info("Applying default settings...")
self.deviceFileSelector.selectFile(settings.devicesJsonFile())
self.tesseractFileSelector.selectFile(settings.tesseractPath())
self.deviceComboBox.selectDevice(settings.deviceUuid())
self.knnModelSelector.selectFile(settings.knnModelFile())
# vself.siftDatabaseSelector.selectFile(settings.siftDatabaseFile())
self.ocrQueueModel = OcrQueueModel(self)
self.ocrQueue.setModel(self.ocrQueueModel)
@ -94,10 +90,6 @@ class TabOcr_Device(Ui_TabOcr_Device, QWidget):
if selectedFiles := self.knnModelSelector.selectedFiles():
self.knnModel = cv2.ml.KNearest.load(selectedFiles[0])
def siftDatabaseFileSelected(self):
if selectedFiles := self.siftDatabaseSelector.selectedFiles():
self.siftDatabase = SIFTDatabase(selectedFiles[0])
def phashDatabaseFileSelected(self):
if selectedFiles := self.phashDatabaseSelector.selectedFiles():
self.phashDatabase = ImagePHashDatabase(selectedFiles[0])

View File

@ -477,10 +477,12 @@ validation</translation>
<location filename="../../implements/settings/settingsAndreal.py" line="82"/>
<location filename="../../implements/settings/settingsAndreal.py" line="85"/>
<location filename="../../implements/settings/settingsGeneral.py" line="107"/>
<location filename="../../implements/settings/settingsOcr.py" line="137"/>
<location filename="../../implements/settings/settingsOcr.py" line="140"/>
<location filename="../../implements/settings/settingsOcr.py" line="143"/>
<location filename="../../implements/settings/settingsOcr.py" line="146"/>
<location filename="../../implements/settings/settingsOcr.py" line="190"/>
<location filename="../../implements/settings/settingsOcr.py" line="193"/>
<location filename="../../implements/settings/settingsOcr.py" line="196"/>
<location filename="../../implements/settings/settingsOcr.py" line="199"/>
<location filename="../../implements/settings/settingsOcr.py" line="202"/>
<location filename="../../implements/settings/settingsOcr.py" line="205"/>
<source>resetButton</source>
<translation>Reset</translation>
</message>
@ -515,30 +517,40 @@ validation</translation>
<translation>Database URL</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="134"/>
<location filename="../../implements/settings/settingsOcr.py" line="187"/>
<source>ocr.title</source>
<translation>OCR</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="136"/>
<location filename="../../implements/settings/settingsOcr.py" line="189"/>
<source>ocr.devicesJson.label</source>
<translation>Default devices.json</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="139"/>
<location filename="../../implements/settings/settingsOcr.py" line="192"/>
<source>ocr.deviceUuid.label</source>
<translation>Default device</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="142"/>
<location filename="../../implements/settings/settingsOcr.py" line="195"/>
<source>ocr.knnModelFile.label</source>
<translation>Default KNearest model</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="145"/>
<location filename="../../implements/settings/settingsOcr.py" line="198"/>
<source>ocr.b30KnnModelFile.label</source>
<translation>Default B30 KNearest model</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="201"/>
<source>ocr.siftDatabaseFile.label</source>
<translation>Default SIFT database file</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="204"/>
<source>ocr.phashDatabaseFile.label</source>
<translation>Default image PHash database</translation>
</message>
</context>
<context>
<name>SongIdSelector</name>
@ -570,7 +582,7 @@ validation</translation>
<message>
<location filename="../../designer/tabs/tabAbout.ui" line="86"/>
<source>Version Info</source>
<translation type="unfinished"></translation>
<translation>Version Info</translation>
</message>
</context>
<context>
@ -704,8 +716,8 @@ validation</translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_B30.ui" line="62"/>
<source>siftDatabaseSelector.title</source>
<translation>Select Image SIFT Database</translation>
<source>phashDatabaseSelector.title</source>
<translation>Select Image PHash Database</translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_B30.ui" line="74"/>
@ -726,32 +738,31 @@ validation</translation>
<translation>Select Device</translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="33"/>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="35"/>
<source>deviceSelector.useAutoFactor</source>
<translation>Auto calculate factor</translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="70"/>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="81"/>
<source>knnModelSelector.title</source>
<translation>Select KNearest Model</translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="101"/>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="112"/>
<source>tesseractSelector.title</source>
<translation>Select tesseract Path</translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="129"/>
<source>siftDatabaseSelector.title</source>
<translation>Select Image SIFT Database</translation>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="140"/>
<source>phashDatabaseSelector.title</source>
<translation>Select Image PHash Database</translation>
</message>
</context>
<context>
<name>TabOverview</name>
<message>
<location filename="../../designer/tabs/tabOverview.ui" line="23"/>
<location filename="../../implements/tabs/tabOverview.py" line="43"/>
<source>databaseDescribeLabel {} {} {} {} {} {}</source>
<extracomment>This database now have {} packs, {} songs, {} difficulties, {} chart info ({} complete) and {} scores.</extracomment>
<translation>There are {} packs, {} songs, {} difficulties, {} chart info ({} complete) and {} scores in database.</translation>
</message>
</context>
@ -841,7 +852,7 @@ validation</translation>
<translation>Generate</translation>
</message>
<message>
<location filename="../../implements/tabs/tabTools/tabTools_Andreal.py" line="82"/>
<location filename="../../implements/tabs/tabTools/tabTools_Andreal.py" line="138"/>
<source>imageWhatIsThisDialog.description</source>
<translation>Generate image of...&lt;ul&gt;&lt;li&gt;/a - the most recent score&lt;/li&gt;&lt;li&gt;/a b30 - best30 image&lt;/li&gt;&lt;li&gt;/a info - best score of the selected chart&lt;/li&gt;&lt;/ul&gt;</translation>
</message>

View File

@ -476,10 +476,12 @@
<location filename="../../implements/settings/settingsAndreal.py" line="82"/>
<location filename="../../implements/settings/settingsAndreal.py" line="85"/>
<location filename="../../implements/settings/settingsGeneral.py" line="107"/>
<location filename="../../implements/settings/settingsOcr.py" line="137"/>
<location filename="../../implements/settings/settingsOcr.py" line="140"/>
<location filename="../../implements/settings/settingsOcr.py" line="143"/>
<location filename="../../implements/settings/settingsOcr.py" line="146"/>
<location filename="../../implements/settings/settingsOcr.py" line="190"/>
<location filename="../../implements/settings/settingsOcr.py" line="193"/>
<location filename="../../implements/settings/settingsOcr.py" line="196"/>
<location filename="../../implements/settings/settingsOcr.py" line="199"/>
<location filename="../../implements/settings/settingsOcr.py" line="202"/>
<location filename="../../implements/settings/settingsOcr.py" line="205"/>
<source>resetButton</source>
<translation></translation>
</message>
@ -514,30 +516,40 @@
<translation> URL</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="134"/>
<location filename="../../implements/settings/settingsOcr.py" line="187"/>
<source>ocr.title</source>
<translation>OCR</translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="136"/>
<location filename="../../implements/settings/settingsOcr.py" line="189"/>
<source>ocr.devicesJson.label</source>
<translation></translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="139"/>
<location filename="../../implements/settings/settingsOcr.py" line="192"/>
<source>ocr.deviceUuid.label</source>
<translation></translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="142"/>
<location filename="../../implements/settings/settingsOcr.py" line="195"/>
<source>ocr.knnModelFile.label</source>
<translation> KNearest </translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="145"/>
<location filename="../../implements/settings/settingsOcr.py" line="198"/>
<source>ocr.b30KnnModelFile.label</source>
<translation> B30 KNearest </translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="201"/>
<source>ocr.siftDatabaseFile.label</source>
<translation> SIFT </translation>
</message>
<message>
<location filename="../../implements/settings/settingsOcr.py" line="204"/>
<source>ocr.phashDatabaseFile.label</source>
<translation> PHash </translation>
</message>
</context>
<context>
<name>SongIdSelector</name>
@ -569,7 +581,7 @@
<message>
<location filename="../../designer/tabs/tabAbout.ui" line="86"/>
<source>Version Info</source>
<translation type="unfinished"></translation>
<translation></translation>
</message>
</context>
<context>
@ -703,8 +715,8 @@
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_B30.ui" line="62"/>
<source>siftDatabaseSelector.title</source>
<translation> SIFT </translation>
<source>phashDatabaseSelector.title</source>
<translation> PHash </translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_B30.ui" line="74"/>
@ -725,32 +737,31 @@
<translation></translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="33"/>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="35"/>
<source>deviceSelector.useAutoFactor</source>
<translation> factor</translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="70"/>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="81"/>
<source>knnModelSelector.title</source>
<translation> KNearest </translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="101"/>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="112"/>
<source>tesseractSelector.title</source>
<translation> tesseract </translation>
</message>
<message>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="129"/>
<source>siftDatabaseSelector.title</source>
<translation> SIFT </translation>
<location filename="../../designer/tabs/tabOcr/tabOcr_Device.ui" line="140"/>
<source>phashDatabaseSelector.title</source>
<translation> PHash </translation>
</message>
</context>
<context>
<name>TabOverview</name>
<message>
<location filename="../../designer/tabs/tabOverview.ui" line="23"/>
<location filename="../../implements/tabs/tabOverview.py" line="43"/>
<source>databaseDescribeLabel {} {} {} {} {} {}</source>
<extracomment>This database now have {} packs, {} songs, {} difficulties, {} chart info ({} complete) and {} scores.</extracomment>
<translation> {} {} {} {} {} {} </translation>
</message>
</context>
@ -840,7 +851,7 @@
<translation></translation>
</message>
<message>
<location filename="../../implements/tabs/tabTools/tabTools_Andreal.py" line="82"/>
<location filename="../../implements/tabs/tabTools/tabTools_Andreal.py" line="138"/>
<source>imageWhatIsThisDialog.description</source>
<translation>&lt;ul&gt;&lt;li&gt;/a - &lt;/li&gt;&lt;li&gt;/a b30 - /a b30&lt;/li&gt;&lt;li&gt;/a info - &lt;/li&gt;&lt;/ul&gt;</translation>
</message>