pyhook free

README.md

@@ -14,23 +14,7 @@
 conda create -n ysfish python=3.6
 conda activate ysfish
 ```
-安装<font color=#66CCFF>**python3.7或以下**</font>版本，否则pyHook要自己编译。
-
-## pyHook安装
-安装PyUserInput前需要安装<font color=#66CCFF>**pyHook**</font>适用于win10的版本可以在 [这里](https://www.lfd.uci.edu/~gohlke/pythonlibs/) 找到。
-
-适用于python3.6的文件为：\
-64位：[pyHook‑1.5.1‑cp36‑cp36m‑win_amd64.whl](https://download.lfd.uci.edu/pythonlibs/y2rycu7g/cp36/pyHook-1.5.1-cp36-cp36m-win_amd64.whl) \
-32位：[pyHook‑1.5.1‑cp36‑cp36m‑win32.whl](https://download.lfd.uci.edu/pythonlibs/y2rycu7g/cp36/pyHook-1.5.1-cp36-cp36m-win32.whl)
-
-下载文件后将命令行切换到文件所在目录，执行:
-```shell
-pip install [文件名].whl
-```
-例如在64位的python3.6下安装：
-```shell
-pip install pyHook-1.5.1-cp36-cp36m-win_amd64.whl
-```
+推荐安装<font color=#66CCFF>**python3.7或以下**</font>版本。
 
 ## 下载工程代码
 使用git下载，[git安装教程](https://www.cnblogs.com/xiaoliu66/p/9404963.html):
@@ -46,7 +30,7 @@ cd genshin_auto_fish
 ```
 执行以下命令安装依赖:
 ```shell
-pip install -U pip
+python -m pip install -U pip
 python requirements.py
 ```
 如果要使用显卡进行加速需要 [安装CUDA和cudnn](https://zhuanlan.zhihu.com/p/94220564?utm_source=wechat_session&ivk_sa=1024320u) 安装后无视上面的命令用下面这条安装gpu版:
@@ -68,17 +52,6 @@ python setup.py develop
 下载预训练[权重](https://github.com/7eu7d7/genshin_auto_fish/releases/tag/weights) (.pth文件),
 下载后将权重文件放在 <font color=#66CCFF>**工程目录/weights**</font> 下
 
-## 可能的问题
-如果后续运行出现<font color=red>ModuleNotFoundError: No module named 'windows'</font>\
-把 [anaconda安装路径]/envs/ysfish/Lib/site-packages/pymouse/\_\_init\_\_.py中的
-```python
-from windows import PyMouse, PyMouseEvent
-```
-改为
-```python
-from .windows import PyMouse, PyMouseEvent
-```
-
 # YOLOX训练工作流程
 YOLOX部分因为打标签太累所以用半监督学习。标注少量样本后训练模型生成其余样本伪标签再人工修正，不断迭代提高精度。
 样本量较少所以使用迁移学习，在COCO预训练的模型上进行fine-tuning.

fisher/environment.py

@@ -1,9 +1,7 @@
 import numpy as np
 from utils import *
 import cv2
-from pymouse import *
 import pyautogui
-import win32api, win32con
 import time
 from copy import deepcopy
 from collections import Counter
@@ -23,6 +21,7 @@ def __init__(self, predictor, show_det=True):
         self.dist_dict={'hua jiang':130, 'ji yu':80, 'die yu':80, 'jia long':80, 'pao yu':80}
         self.food_rgn=[580,400,740,220]
         self.last_fish_type='hua jiang'
+        #self.last_fish_type='die yu' # 钓雷鸣仙
         self.show_det=show_det
         os.makedirs('img_tmp/', exist_ok=True)
 
@@ -32,22 +31,18 @@ def get_fish_types(self, n=12, rate=0.6):
         for i in range(n):
             obj_list = self.predictor.image_det(cap())
             if obj_list is None:
-                win32api.mouse_event(win32con.MOUSEEVENTF_MOVE, 70 * fx(i), 0, 0, 0)
+                mouse_move(70 * fx(i), 0)
                 time.sleep(0.1)
                 continue
             cls_list = set([x[0] for x in obj_list])
             counter.update(cls_list)
-            win32api.mouse_event(win32con.MOUSEEVENTF_MOVE, 70 * fx(i), 0, 0, 0)
+            mouse_move(70 * fx(i), 0)
             time.sleep(0.2)
-            # pyautogui.moveRel(50, 0, duration=0.5)
-            # for u in range(1,51):
-            #    mosue.move(sx + u, sy)
-            #    time.sleep(0.003)
         fish_list = [k for k, v in dict(counter).items() if v / n >= rate]
         return fish_list
 
     def throw_rod(self, fish_type):
-        win32api.mouse_event(win32con.MOUSEEVENTF_LEFTDOWN, 0, 0)
+        mouse_down(960, 540)
         time.sleep(1)
 
         def move_func(dist):
@@ -64,7 +59,7 @@ def move_func(dist):
 
                 rod_info = sorted(list(filter(lambda x: x[0] == 'rod', obj_list)), key=lambda x: x[1], reverse=True)
                 if len(rod_info)<=0:
-                    win32api.mouse_event(win32con.MOUSEEVENTF_MOVE, np.random.randint(-50,50), np.random.randint(-50,50), 0, 0)
+                    mouse_move(np.random.randint(-50,50), np.random.randint(-50,50))
                     time.sleep(0.1)
                     continue
                 rod_info=rod_info[0]
@@ -86,19 +81,18 @@ def move_func(dist):
 
                 dx = int(move_func(x_dist))
                 dy = int(move_func(((fish_info[2][3]) + fish_info[2][1]) / 2 - rod_info[2][3]))
-                #win32api.mouse_event(win32con.MOUSEEVENTF_MOVE, fish_info[2][2] - rod_info[2][2] + 50, (fish_info[2][3] + fish_info[2][1]) / 2 - rod_info[2][3], 0, 0)
-                win32api.mouse_event(win32con.MOUSEEVENTF_MOVE, dx, dy, 0, 0)
+                mouse_move(dx, dy)
             except Exception as e:
                 traceback.print_exc()
             #time.sleep(0.3)
-        win32api.mouse_event(win32con.MOUSEEVENTF_LEFTUP, 0, 0)
+        mouse_up(960, 540)
 
     def select_food(self, fish_type):
         pyautogui.press('f')
         time.sleep(1)
         pyautogui.click(1650, 790, button=pyautogui.SECONDARY)
         time.sleep(0.5)
-        bbox_food = match_img(cap(self.food_rgn), self.food_imgs[self.ff_dict[fish_type]], type=cv2.TM_CCOEFF_NORMED)
+        bbox_food = match_img(cap(self.food_rgn), self.food_imgs[self.ff_dict[fish_type]], type=cv2.TM_CCORR_NORMED)
         pyautogui.click(bbox_food[4]+self.food_rgn[0], bbox_food[5]+self.food_rgn[1])
         time.sleep(0.1)
         pyautogui.click(1183, 756)
@@ -113,7 +107,6 @@ def do_fish(self, fish_init=True):
 
 class Fishing:
     def __init__(self, delay=0.1, max_step=100, show_det=True, predictor=None):
-        self.mosue = PyMouse()
         self.t_l = cv2.imread('./imgs/target_left.png')
         self.t_r = cv2.imread('./imgs/target_right.png')
         self.t_n = cv2.imread('./imgs/target_now.png')
@@ -141,7 +134,7 @@ def reset(self):
         return self.get_state()
 
     def drag(self):
-        self.mosue.click(1630,995)
+        mouse_click(1630,995)
 
     def do_action(self, action):
         if action==1:

fishing.py

@@ -180,7 +180,7 @@ def start_fishing(predictor, agent, bite_timeout=20):
             state, reward, done = env.step(action)
             if done:
                 break
-        time.sleep(1)
+        time.sleep(3)
 
 #python fishing.py image -f yolox/exp/yolox_tiny_fish.py -c weights/best_tiny3.pth --conf 0.25 --nms 0.45 --tsize 640 --device gpu
 if __name__ == "__main__":

requirements.py

@@ -11,7 +11,6 @@
 loguru
 matplotlib
 tabulate
-PyUserInput
 tqdm
 pywin32
 PyAutoGUI
@@ -21,10 +20,11 @@
 pymouse
 numpy==1.19.5
 torch==1.7.0+{"cpu" if args.cuda is None else "cu" + args.cuda} -f https://download.pytorch.org/whl/torch_stable.html
---no-deps torchvision==0.8.1+{"cpu" if args.cuda is None else "cu" + args.cuda} -f https://download.pytorch.org/whl/torch_stable.html
-thop
+torchvision==0.8.1+{"cpu" if args.cuda is None else "cu" + args.cuda} --no-deps -f https://download.pytorch.org/whl/torch_stable.html
+thop --no-deps
 git+https://github.com/philferriere/cocoapi.git#subdirectory=PythonAPI
 '''
 
 for line in pkgs.split('\n'):
-    pip.main(['install', *line.split()])
+    if len(line)>0:
+        pip.main(['install', *line.split()])

utils/utils.py

@@ -1,11 +1,25 @@
 import cv2
 import pyautogui
 import numpy as np
+import win32api, win32con
 
 def cap(region=None):
     img = pyautogui.screenshot(region=region) if region else pyautogui.screenshot()
     return cv2.cvtColor(np.asarray(img), cv2.COLOR_RGB2BGR)
 
+def mouse_down(x, y):
+    win32api.mouse_event(win32con.MOUSEEVENTF_LEFTDOWN, x, y, 0, 0)
+
+def mouse_move(dx, dy):
+    win32api.mouse_event(win32con.MOUSEEVENTF_MOVE, dx, dy, 0, 0)
+
+def mouse_up(x, y):
+    win32api.mouse_event(win32con.MOUSEEVENTF_LEFTUP, x, y, 0, 0)
+
+def mouse_click(x, y):
+    mouse_down(x, y)
+    mouse_up(x, y)
+
 def match_img(img, target, type=cv2.TM_CCOEFF):
     h, w = target.shape[:2]
     res = cv2.matchTemplate(img, target, type)