Update v1.3

- Fixed Timeout Retrying - Faster Detector Model Loading - Improved ReadMe - Updated Github Workflow to PreInstall Huggingface Models - Updated Requirements
Vinyzu · Feb 12, 2024 · b33490f · b33490f
1 parent 7949c5f
commit b33490f
Show file tree

Hide file tree

Showing 9 changed files with 85 additions and 44 deletions.
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
@@ -49,5 +49,10 @@ jobs:
       uses: browser-actions/setup-chrome@v1
     - name: Install Chromium Driver
       run: python -m playwright install chromium
+    - name: Install HuggingFace Models
+      run: |
+        pip install -U "huggingface_hub[cli]"
+        huggingface-cli download flavour/CLIP-ViT-B-16-DataComp.XL-s13B-b90K config.json
+        huggingface-cli download CIDAS/clipseg-rd64-refined config.json
     - name: Test with PyTest
       run: pytest
diff --git a/README.md b/README.md
@@ -1,5 +1,6 @@
-# reCognizer v1.2
-![Tests & Linting](https://github.com/Vinyzu/recognizer/actions/workflows/tests.yml/badge.svg)
+# reCognizer v1.3
+![Tests & Linting](https://github.com/Vinyzu/recognizer/actions/workflows/ci.yml/badge.svg)
+
 
 #### reCognizer is a free-to-use AI based [reCaptcha](https://developers.google.com/recaptcha) Solver. <br> Usable with an easy-to-use API, also available for Async and Sync Playwright. <br> You can pass almost any format into the Challenger, from full-page screenshots, only-captcha images and no-border images to single images in a list.
 
@@ -13,6 +14,16 @@ pip install recognizer
 
 ---
 
+## Examples
+
+### Possible Image Inputs
+![Accepted Formats](https://i.ibb.co/nztTD9Z/formats.png)
+
+### Example Solve Video (Good IP & Botright)
+https://github.com/Vinyzu/recognizer/assets/50874994/95a713e3-bb46-474b-994f-cb3dacae9279
+
+---
+
 ## Basic Usage
 
 ```py
@@ -107,7 +118,7 @@ asyncio.run(main())
 
 ---
 
-![Version](https://img.shields.io/badge/reCognizer-v1.2-blue)
+![Version](https://img.shields.io/badge/reCognizer-v1.3-blue)
 ![License](https://img.shields.io/badge/License-GNU%20GPL-green)
 ![Python](https://img.shields.io/badge/Python-v3.x-lightgrey)
 

diff --git a/recognizer/__init__.py b/recognizer/__init__.py
@@ -2,6 +2,6 @@
 
 from .components.detector import Detector
 
-VERSION = 1.2
+VERSION = 1.3
 
 __all__ = ["Detector", "VERSION"]
diff --git a/recognizer/agents/playwright/async_control.py b/recognizer/agents/playwright/async_control.py
@@ -31,11 +31,6 @@ def __init__(self, page: Page, click_timeout: Optional[int] = None, retry_times:
         self.dynamic: bool = False
         self.captcha_token: Optional[str] = None
 
-    def check_retry(self):
-        self.retried += 1
-        if self.retried >= self.retry_times:
-            raise RecursionError(f"Exceeded maximum retry times of {self.retry_times}")
-
     async def route_handler(self, route: Route, request: Request) -> None:
         response = await route.fetch()
         await route.fulfill(response=response)  # Instant Fulfillment to save Time
@@ -69,10 +64,11 @@ async def check_captcha_visible(self):
         label_obj = captcha_frame.locator("//strong")
         try:
             await label_obj.wait_for(state="visible", timeout=10000)
-            return True
         except TimeoutError:
             return False
 
+        return await label_obj.is_visible()
+
     async def click_checkbox(self) -> bool:
         # Clicking Captcha Checkbox
         try:
@@ -98,7 +94,9 @@ async def detect_tiles(self, prompt: str, area_captcha: bool) -> bool:
 
     async def load_captcha(self, captcha_frame: Optional[FrameLocator] = None, reset: Optional[bool] = False) -> Union[str, bool]:
         # Retrying
-        self.check_retry()
+        self.retried += 1
+        if self.retried >= self.retry_times:
+            raise RecursionError(f"Exceeded maximum retry times of {self.retry_times}")
 
         if not await self.check_captcha_visible():
             if captcha_token := await self.check_result():
@@ -111,8 +109,11 @@ async def load_captcha(self, captcha_frame: Optional[FrameLocator] = None, reset
         # Clicking Reload Button
         if reset:
             assert isinstance(captcha_frame, FrameLocator)
-            reload_button = captcha_frame.locator("#recaptcha-verify-button")
-            await reload_button.click()
+            try:
+                reload_button = captcha_frame.locator("#recaptcha-reload-button")
+                await reload_button.click()
+            except TimeoutError:
+                return await self.load_captcha()
 
             # Resetting Values
             self.dynamic = False
@@ -147,7 +148,7 @@ async def handle_recaptcha(self) -> Union[str, bool]:
         area_captcha = len(recaptcha_tiles) == 16
         result_clicked = await self.detect_tiles(prompt, area_captcha)
 
-        if self.dynamic and not len(recaptcha_tiles) == 16:
+        if self.dynamic and not area_captcha:
             while result_clicked:
                 await self.page.wait_for_timeout(5000)
                 result_clicked = await self.detect_tiles(prompt, area_captcha)
@@ -170,9 +171,13 @@ async def handle_recaptcha(self) -> Union[str, bool]:
 
             await self.page.wait_for_timeout(1000)
 
+        # Check if error occurred whilst solving
+        incorrect = self.page.locator("[class='rc-imageselect-incorrect-response']")
+        errors = self.page.locator("[class *= 'rc-imageselect-error']")
+        if await incorrect.is_visible() or any([await error.is_visible() for error in await errors.all()]):
+            await self.load_captcha(captcha_frame, reset=True)
+
         # Retrying
-        self.check_retry()
-        await self.load_captcha(captcha_frame, reset=True)
         return await self.handle_recaptcha()
 
     async def solve_recaptcha(self) -> Union[str, bool]:

diff --git a/recognizer/agents/playwright/sync_control.py b/recognizer/agents/playwright/sync_control.py
@@ -31,11 +31,6 @@ def __init__(self, page: Page, click_timeout: Optional[int] = None, retry_times:
         self.dynamic: bool = False
         self.captcha_token: Optional[str] = None
 
-    def check_retry(self):
-        self.retried += 1
-        if self.retried >= self.retry_times:
-            raise RecursionError(f"Exceeded maximum retry times of {self.retry_times}")
-
     def route_handler(self, route: Route, request: Request) -> None:
         response = route.fetch()
         route.fulfill(response=response)  # Instant Fulfillment to save Time
@@ -69,10 +64,11 @@ def check_captcha_visible(self):
         label_obj = captcha_frame.locator("//strong")
         try:
             label_obj.wait_for(state="visible", timeout=10000)
-            return True
         except TimeoutError:
             return False
 
+        return label_obj.is_visible()
+
     def click_checkbox(self) -> bool:
         # Clicking Captcha Checkbox
         try:
@@ -98,7 +94,9 @@ def detect_tiles(self, prompt: str, area_captcha: bool) -> bool:
 
     def load_captcha(self, captcha_frame: Optional[FrameLocator] = None, reset: Optional[bool] = False) -> Union[str, bool]:
         # Retrying
-        self.check_retry()
+        self.retried += 1
+        if self.retried >= self.retry_times:
+            raise RecursionError(f"Exceeded maximum retry times of {self.retry_times}")
 
         if not self.check_captcha_visible():
             if captcha_token := self.check_result():
@@ -111,8 +109,11 @@ def load_captcha(self, captcha_frame: Optional[FrameLocator] = None, reset: Opti
         # Clicking Reload Button
         if reset:
             assert isinstance(captcha_frame, FrameLocator)
-            reload_button = captcha_frame.locator("#recaptcha-verify-button")
-            reload_button.click()
+            try:
+                reload_button = captcha_frame.locator("#recaptcha-reload-button")
+                reload_button.click()
+            except TimeoutError:
+                return self.load_captcha()
 
             # Resetting Values
             self.dynamic = False
@@ -147,7 +148,7 @@ def handle_recaptcha(self) -> Union[str, bool]:
         area_captcha = len(recaptcha_tiles) == 16
         result_clicked = self.detect_tiles(prompt, area_captcha)
 
-        if self.dynamic and not len(recaptcha_tiles) == 16:
+        if self.dynamic and not area_captcha:
             while result_clicked:
                 self.page.wait_for_timeout(5000)
                 result_clicked = self.detect_tiles(prompt, area_captcha)
@@ -170,9 +171,13 @@ def handle_recaptcha(self) -> Union[str, bool]:
 
             self.page.wait_for_timeout(1000)
 
+        # Check if error occurred whilst solving
+        incorrect = self.page.locator("[class='rc-imageselect-incorrect-response']")
+        errors = self.page.locator("[class *= 'rc-imageselect-error']")
+        if incorrect.is_visible() or any([error.is_visible() for error in errors.all()]):
+            self.load_captcha(captcha_frame, reset=True)
+
         # Retrying
-        self.check_retry()
-        self.load_captcha(captcha_frame, reset=True)
         return self.handle_recaptcha()
 
     def solve_recaptcha(self) -> Union[str, bool]:

diff --git a/recognizer/components/detector.py b/recognizer/components/detector.py
@@ -3,10 +3,10 @@
 import math
 import random
 import warnings
-from concurrent.futures import ThreadPoolExecutor
+from concurrent.futures import Future, ThreadPoolExecutor
 from os import PathLike
 from pathlib import Path
-from typing import List, Optional, Sequence, Tuple, Union
+from typing import Callable, List, Optional, Sequence, Tuple, Union
 
 import cv2
 from numpy import generic, uint8
@@ -23,12 +23,16 @@
 class DetectionModels:
     def __init__(self) -> None:
         # Preloading: Loading Models takes ~9 seconds
-        set_num_threads(2)
-        self.executor = ThreadPoolExecutor(max_workers=2)
+        set_num_threads(5)
+        self.executor = ThreadPoolExecutor(max_workers=5)
+        self.loading_futures: List[Future[Callable[..., None]]] = []
 
         try:
-            self.yolo_loading_feature = self.executor.submit(self._load_yolo_detector)
-            self.clip_loading_feature = self.executor.submit(self._load_clip_detector)
+            self.loading_futures.append(self.executor.submit(self._load_yolo_detector))
+            self.loading_futures.append(self.executor.submit(self._load_vit_model))
+            self.loading_futures.append(self.executor.submit(self._load_vit_processor))
+            self.loading_futures.append(self.executor.submit(self._load_seg_model))
+            self.loading_futures.append(self.executor.submit(self._load_seg_processor))
         except Exception as e:
             self.executor.shutdown(wait=True, cancel_futures=True)
             raise e
@@ -38,20 +42,31 @@ def _load_yolo_detector(self):
 
         self.yolo_model = YOLO("yolov8m-seg.pt")
 
-    def _load_clip_detector(self):
-        from transformers import CLIPModel, CLIPProcessor, CLIPSegForImageSegmentation, CLIPSegProcessor
+    def _load_vit_model(self):
+        from transformers import CLIPModel
 
         self.vit_model = CLIPModel.from_pretrained("flavour/CLIP-ViT-B-16-DataComp.XL-s13B-b90K")
+
+    def _load_vit_processor(self):
+        from transformers import CLIPProcessor
+
         self.vit_processor = CLIPProcessor.from_pretrained("flavour/CLIP-ViT-B-16-DataComp.XL-s13B-b90K")
 
+    def _load_seg_model(self):
+        from transformers import CLIPSegForImageSegmentation
+
         self.seg_model = CLIPSegForImageSegmentation.from_pretrained("CIDAS/clipseg-rd64-refined")
+
+    def _load_seg_processor(self):
+        from transformers import CLIPSegProcessor
+
         self.seg_processor = CLIPSegProcessor.from_pretrained("CIDAS/clipseg-rd64-refined")
 
     def check_loaded(self):
         try:
-            if not self.yolo_loading_feature.done() or not self.clip_loading_feature.done():
-                self.yolo_loading_feature.result()
-                self.clip_loading_feature.result()
+            if not all([future.done() for future in self.loading_futures]):
+                for future in self.loading_futures:
+                    future.result()
 
             assert self.yolo_model
             assert self.seg_model

diff --git a/requirements-test.txt b/requirements-test.txt
@@ -2,7 +2,7 @@
 flake8==7.0.0
 tox==4.12.1
 pytest==7.4.4
-pytest_asyncio==0.23.4
+pytest_asyncio==0.23.5
 mypy==1.8.0
 types-setuptools==69.0.0.20240125
 botright==0.5

diff --git a/requirements.txt b/requirements.txt
@@ -1,7 +1,7 @@
-setuptools~=69.0.3
+setuptools~=69.1.0
 opencv-python~=4.9.0.80
 imageio~=2.33.1
 ultralytics~=8.1.9
 transformers~=4.37.2
-numpy~=1.26.3
-playwright~=1.41.0
+numpy~=1.26.4
+playwright~=1.41.2
diff --git a/setup.cfg b/setup.cfg
@@ -1,7 +1,7 @@
 [metadata]
 name = recognizer
 version = attr: recognizer.VERSION
-description = 🦉 Gracefully face reCAPTCHA challenge with ultralytics YOLOv8-seg and CLIPs VIT-B/32. Implemented in playwright or an easy-to-use API.
+description = 🦉Gracefully face reCAPTCHA challenge with ultralytics YOLOv8-seg, CLIPs VIT-B/16 and CLIP-Seg/RD64. Implemented in playwright or an easy-to-use API.
 long_description = file: README.md
 long_description_content_type = text/markdown
 author = Vinyzu