    : , -   
 


     : , -         ,      .  20     :       vision-  OpenCV  YOLO, reinforcement learning      (VLM/LLM).

 ,   ,    ,  ,      .      ,  , ,    open-source  (SIMA, YOLO-aim, MAA, Mineflayer  .).

   ,    ,   ,     ,         .





 

    : , -   








 1.     



        ,  ,  ,     . Ÿ      (, , ),    ,            .



          .     ,    ,   .        :   ( ,   ), ,   ,     .        DeepMind, OpenAI, Meta AI             .



1.1.    



    ,           .      :



         ,     .

      ,  ,    ,     .

       , ,     ,   .

     ,  ,    ,   .

      ,  ,    perception  control.



      .           :   , wallhack, speedhack,     . Vision-based          ,     ,      .              ,        .



1.2.    



     ,          .



-,   .    -. Atari 2600, Doom, StarCraft II, Dota 2, Minecraft, NetHack       .  AlphaGo, OpenAI Five  AlphaStar  ,          .        : ,       ,     .     SIMA (Scalable Instructable Multiworld Agent)  DeepMind.



-,     .  AAA-         .      .       ,   edge-cases, ,      ,       .



-, .            ,           .                 .



-,  .  MMO, idle-, gacha   live-service        .       ,    . ,              .



-, .          ,      ,  reinforcement learning    .     :     ,  .   perception      .



1.3.    



       .          .



   .   .            . : AutoHotkey, AutoIt, pyautogui, Keyboard Maestro.        ,     .        (  ,  ,    single-player)             .



Memory-based .         .      HP, ,  ,    -  .   reverse engineering:  ,  -,    (   Cheat Engine  ).     .      ,     ,    .



Network / Packet bots.         .    MMO.    -  ,     server-side .



Vision-based .          .     ,       .           ,        .       ,           .             ,   UI.



API-based  SDK-based .       . : python-sc2  pysc2  StarCraft II, Mineflayer  Minecraft,  -API    .    ,   :           ,    perception.



Reinforcement Learning  end-to-end .             ,    .       .     ,     , Dota 2  StarCraft II.



 .        .         9598 %   .         ,      ,          .



1.4.   vision-



 vision-based      :



1.   (screen capture)            .

2.          ,  ROI, .

3.  (perception)   ,    , OCR,    .

4.   (state representation)     CV   :   ,  HP,   ,   .

5.   (decision making)  state machine, behaviour tree,     LLM.

6.   (action planning)        .

7.  (execution)           .

8.       ,  ,     .



               OpenCV      GPU-.        .



1.5.    



   ,       (  Python)           .        ,   .              .



     vision-based  RL-,      . Memory-hacking       :         ,        -.



       :                       .



 2.     



       ,     .                .



2.1.   (19501970-)



 1951            Ferranti Mark 1   .      OXO  -  EDSAC.          .             ,      .    Turochamp, , ,            .



        .       -   Deep Blue,     1997 ,   .  : Deep Blue            ,         .       ,   .



2.2.      (19701990-)



          ,    ,     ,  .    Pac-Man (1980).    Blinky, Pinky, Inky  Clyde      . Blinky    . Pinky           . Inky    ,     Blinky,  . Clyde       .             ,     .



 1990-      . Quake, Unreal Tournament, Half-Life    .    Omni-bot,    -   .       ,    ,  ,          .  20082012    2K BotPrize       .    ,  -           Unreal Tournament 2004.  2012       .



2.3.   MMO  - (20002010-)



   World of Warcraft, Lineage, Runescape   MMO     .        ,     .   :      ,      .



       . Warden  World of Warcraft,  BattlEye, Easy Anti-Cheat, PunkBuster, Valve Anti-Cheat.    :    ,   ,    .      DLL-,  API,              .



2.4.    (20132019)



   20132015 . DeepMind    DQN   Q-,       Atari,        .      ,  end-to-end          ,      .



 2016  AlphaGo       ,        -      .   OpenAI Five (Dota 2, 2018)  AlphaStar (StarCraft II, 2019).        ,      ,   .



   : OpenAI Gym ( Gymnasium), ViZDoom, Unity ML-Agents, MineRL, NetHack Learning Environment.            .



2.5.   (20202026)



 2020      .



      . DeepMind  SIMA  ,          3D-,     ,     . Microsoft  Muse  ,       .    ,     .



    vision-based  LLM-powered .      ( YOLO)    (GPT-4o, Claude, Gemini, LLaVA, Qwen-VL)    ,            .   ,        ,        .



    agentic AI   :  NPC,   ,         .      ,   .



2.6.  



    :



     , , template matching       .   ,      .

  end-to-end         ,     .

               .

            .



         ,  ,     ,    .



 3.  :   



    

  -   



   20 

2026







 1.     

 2.     

 3.  :   

 4. ,    

 5.     

 6.     

 7.  :  Template Matching  YOLO

 8. OCR    

 9.  : , , 

 10.    Python + OpenCV

 11.       

 12. Reinforcement Learning   

 13.    (ViZDoom, ML-Agents, Gym)

 14. -  FPS, MMO  

 15. -   

 16. Vision-Language Models  LLM-  

 17. -:    

 18. ,   

 19.  -  

 20.    



 1.     



        ,  ,     .     (),    ,      .



  ?



             (DeepMind AlphaGo, OpenAI Five, SIMA).

        ,      .

          .

               RL.

     MMO  idle-       (,  ).



 



       :



1.  /        (AutoHotkey, AutoIt, pyautogui).

2. Memory-based        ( reverse engineering).

3. Vision-based        (OpenCV, YOLO, ).

4. API / SDK-based        (python-sc2  StarCraft II, Mineflayer  Minecraft).

5. Reinforcement Learning / End-to-end          .



        vision-based  RL-,       (Vision-Language Models),          .



 2.     



       .   1950-          (Nim, OXO).



 



 19511952  Christopher Strachey (Nim), Alexander Douglas (OXO)        .

 197080-    (Space Invaders, Pac-Man).   Pac-Man         (FSM).

 1990-   FPS- (Quake, Unreal Tournament).  framework'  Omni-bot. -  BotPrize (Turing Test   ).

 20002010-   -  MMO (World of Warcraft, Lineage).   (Warden, BattlEye, Easy Anti-Cheat).

 20132016  DeepMind: DQN  Atari, AlphaGo. OpenAI Gym.  ViZDoom.

 20182019  OpenAI Five (Dota 2), AlphaStar (StarCraft II)  ,   .

 20242026  SIMA (DeepMind), Vision-Language Models, LLM-,         .   agentic   .



      (rule-based)     , ,   foundation-,     .



 3.  :   



   



 Memory bots     ( HP, , ).   ,       reverse engineering.

 Packet / Network bots      .     MMO.

 Vision bots      .        (  ).

 Hybrid   .



  



 Clicker / Macro      .

 State-machine bots   :     ,  HP   .

 Classical CV bots  template matching,  ,  + .

 Deep Learning bots  YOLO/  +  + .

 RL agents   end-to-end (PPO, DQN, A3C  ..).

 LLM / VLM agents                .



 4. ,    



       (ToS)  -.      ,         (       ).



 



       .

      (,  ).

       (gold farming).

  :    , ,  .



          .            -.



 5.     



  (Computer Vision)     vision-based .          (),      :  , HP, , ,  .



  CV   



 Object Detection        bounding box' (, NPC, , ).

 Classification  ,     (, , ,  ).

 Segmentation     (   ).

 OCR    (, ,  ,  ).

 Tracking      (  ).

 Template Matching      .



 :   ?  ? / ?   ?   ? .



 6.     



      .     ,     (FPS).



  



 mss (Python)      - .

 PIL / Pillow ImageGrab  ,  .

 dxcam (Windows)      Desktop Duplication API.

 OpenCV VideoCapture        ().

  API (BitBlt, DXGI)   ,    .







 :   (resize),    ,  (GaussianBlur),   (threshold),  ,    (ROI),  ,      (HSV, LAB).



    mss:



import mss

import numpy as np

import cv2



with mss.mss() as sct:

monitor = sct.monitors[1]

img = np.array(sct.grab(monitor))

frame = cv2.cvtColor(img, cv2.COLOR_BGRA2BGR)



 7.  :  Template Matching  YOLO



Template Matching



  OpenCV (cv2.matchTemplate).         .     UI-      .     , ,    .



 



  (HSV threshold),   (findContours),  , HOG + SVM              GPU.



Deep Learning 



 YOLO (v5v12)       . -,  ,     .

 SSD, EfficientDet, RetinaNet  .

 RT-DETR, RF-DETR    .



       ,  (LabelImg, CVAT, Roboflow),   (ultralytics YOLO)    ONNX / TensorRT    .



 8. OCR    



        :  HP,  , ,  ,   .







 Tesseract OCR  ,     .

 EasyOCR, PaddleOCR       OCR.

             (     OCR).



:   ROI  ,  ,  ,    .      template matching    .



 9.  : , , 



     ,   .            .



  



 pyautogui / pydirectinput    .

 pynput     .

 Windows API (SendInput, mouse_event)    .

   /  (Interception, HID)        .

 Arduino / hardware spoofing    /.



     ,     ,          .



 10.    Python + OpenCV



    : Python + OpenCV + mss + pyautogui/pydirectinput.       , idle, puzzle   single-player .



   



1.  .

2.    ( /  / )   ,    .

3.   (template matching  ).

4.   (if-else   state machine).

5.   (,  ).

6.    .



     :   ,  ,  ,   turn-based .  open-source  (  Tetris, fishing bots, clicker-)     .



 11.       



   template matching   (,  ,  ),     .



 



     (YOLO, RT-DETR).

    /  .

  (   ).

   ( aimbot'  ).

    ( , behavioral cloning).



Behavioral Cloning   : ,    ( + ),      .    ,     .



 12. Reinforcement Learning   



   (Reinforcement Learning)   ,      ,   .      RL.



 



 DQN ( : Double DQN, Dueling, Rainbow)  value-based.

 PPO (Proximal Policy Optimization)        policy-gradient .

 A3C / A2C   actor-critic.

 SAC, TD3     .

 AlphaZero / MuZero   +  (   perfect information).



 : sample efficiency (   ), reward shaping,   (partial observability),  (exploration)     .



 13.    (ViZDoom, ML-Agents, Gym)



    RL-   :



 OpenAI Gym / Gymnasium    .

 ViZDoom  Doom    RL   .

 Unity ML-Agents   toolkit     Unity.

 DeepMind Lab, Procgen, MineRL, NetHack Learning Environment    .

 Stable-Baselines3, RLlib, CleanRL, Sample Factory    RL.

 python-sc2, pysc2   StarCraft II.



         +   ,  imitation learning   .



 14. -  FPS, MMO  



FPS



 : YOLO-  ? Kalman filter /  ?   ?    ? .    (  RL-)   .  open-source:  YOLO-aim , BOTTI  .



MMO / RPG



   state machine,  UI, pathfinding  ,    .    CV +  + () imitation learning. :   Genshin Impact, New World,  MAA- .



  RTS



  API- (StarCraft II)     + . AlphaStar ,  RL    .



 15. -   



  (BattlEye, Easy Anti-Cheat, Vanguard, VAC, Ricochet  .)   :



     .

   (  ,  ,   ).

     .

   kernel-level .

    machine learning   .



Vision-based        memory-,        .         offline / single-player /  .



 16. Vision-Language Models  LLM-  



 20232026    :   (GPT-4o, Claude, Gemini, LLaVA, Qwen-VL, DeepSeek-VL  .)         ,    .



 



  VLM-    (   N )        .

       (, OCR, state machine)  9598 % ,  LLM      (, ,  ).

 SIMA-           3D-.



     (     )       pure-LLM .



 17. -:    



 YOLO + aim-   open-source   FPS  real-time   .

 MAA Assistant (Arknights)       vision + script    .

 Mineflayer   API  Minecraft-.

 BOT-MMORPG-AI    imitation learning + computer vision  MMO.

 Peek-a-bot   RL-     Unreal Engine 5.

 DeepMind SIMA  generalist agent,      3D-.

 GamingAgent / LMGame     LLM  grid-based  platformer .



 18. ,   



   ML



OpenCV, ultralytics (YOLO), torchvision, timm, ONNX Runtime, TensorRT, OpenVINO, EasyOCR / PaddleOCR, supervision.



  



mss, dxcam, pyautogui, pydirectinput, pynput, keyboard, mouse.



RL  



Gymnasium, Stable-Baselines3, CleanRL, RLlib, PettingZoo, Unity ML-Agents, ViZDoom.







NumPy, PyTorch / TensorFlow, Roboflow, Label Studio, Weights & Biases, Hydra / OmegaConf.



 19.  -  



  :



      generalist-,        .

  agentic AI     ( NPC,  , procedural storytelling).

  sample efficiency     .

         .

        embodied AI  .

         -.



 20.    



    -       practical-     reinforcement learning.    feedback   .



 



1.    single-player  offline    OpenCV + template matching.

2.   , ROI,     .

3.   YOLO:   ,  ,  FPS.

4.  state machines     .

5.      RL   .

6.     : offline,  , research-only.



          . ,     ,     ,     .



 -------------------------------------------------------------------------------

    .   .

 -------------------------------------------------------------------------------



 3.  :   



    ,   .      .



3.1.      



Memory-based     .   HP, , ID .  ReadProcessMemory  reverse engineering (Cheat Engine, ).     .     ,  ,  .



Packet-based    .        ,   .



Vision-based     .    :    ,    (  ,    ),     .     perception.



Hybrid  . Audio  multimodal  vision ,      .



3.2.   



Clicker/Macro     .

Rule-based  State-machine   HP < 30%  ,  .

Classical CV  template matching, ,  + .

Deep Learning perception + rules  YOLO/  ,  .

Imitation Learning     ;  ,  , compounding errors.

RL agents    ;   .

LLM/VLM agents     /; ,    ,    .



3.3.  



 ;  ; -;    CI.



3.4.  



 ? :  ? YOLO ?  ? if-else/SM ? . , ,   FPS aim   MMO-.

  ? :   (//),  .   UI-heavy .

 : high-level  , low-level .

   : CV/OCR ?   ?   (  LLM).



3.5.  



  ?    < 50100 ?   API?   ?    ?       ?     2020-   :    +   +    .



 4. ,    



  -    ToS.     , HWID-,        RMT.   vision-,  ,  .



   .          ;         RMT      .



 :     PvP;   ; -  .     :  ,  ,  ,   CV  RL.



:     single-player, offline,   research-;    ;     ;     .      .



 5.     



     vision-  . :     ;    ,  ;      GUI.   , , , , .



: object detection (, NPC, , ); classification (//); segmentation; tracking; OCR; template matching;  pose estimation.



:  ?  ? ROI ? perception ?  ?   ? .



 :  ,  UI, , ,  aspect ratio  DPI.  perception?decision        . : precision/recall, mAP, IoU, FPS, latency,    .



    API,     ,    latency   .



 6.     



  :  FPS (3060+),  latency,  /ROI, multi-monitor  DPI,  .



Windows: BitBlt (/), Desktop Duplication API (), mss ( -), dxcam ( ), PIL (/).



:

import mss, numpy as np, cv2

with mss.mss() as sct:

mon = sct.monitors[1]

img = np.array(sct.grab(mon))

frame = cv2.cvtColor(img, cv2.COLOR_BGRA2BGR)



: HWND ? client rect  DPI ? region. Exclusive fullscreen   windowed/borderless.



: resize, BGR/RGB/HSV/gray, ROI, blur, CLAHE, threshold, morphology, inRange. :   ;     ; GPU  ; . :  ,  ,  FPS    API  DPI-awareness.



 : Python, mss  dxcam, OpenCV, numpy.



 7.  :  Template Matching  YOLO



Template Matching (matchTemplate)    UI  ;       .  +  (HSV, inRange, findContours)    . Haar/HOG     .



YOLO (v5v12)   : , , ultralytics,   TensorRT/ONNX. : RT-DETR, EfficientDet.



:   ?  (LabelImg, CVAT, Roboflow) ?  ? fine-tune ?    ?  ? . : NMS, confidence, ,  (ByteTrack, Kalman),  ID.  YOLO    template/  UI. :  ,  ,  ,  .



 8. OCR    



OCR   HP, , , , . : Tesseract, EasyOCR, PaddleOCR;          template .



:  ROI, upscale, gray, , , . : whitelist,  ,   HP.     OCR    .  , ,  ,  fallback.



 9.  : , , 



 : pyautogui/pydirectinput/pynput ? WinAPI ?  HID/ ? Arduino.      ,   .



: / ,   , , ,   sens  raw input. : press/release, , . : vgamepad.



:   ,   180,  ,   .  , ,  InputController.



 10.    Python + OpenCV



: Python + OpenCV + mss + pydirectinput. : capture ? vision ? decisions ? input.  :  ROI, template UI, ,  .



:  ; . State machine   . : overlay , , dry-run.   YOLO/RL/LLM,     .



 11.       



ML: , , , OCR, , behavioral cloning, RL. Behavioral cloning   ,  compounding errors.



: YOLO nano/small, 3002000 , fine-tune, TensorRT.   . :  , FP16/INT8,  .    ,   OpenCV.



 12. Reinforcement Learning   



RL: state, action, reward, policy, value, episode. DQN-; PPO/A2C/SAC; model-based (MuZero). PPO   .



:  , sparse rewards, partial observability.  reward shaping, curriculum, ;  reward hacking. : Gymnasium, ViZDoom, ML-Agents, pysc2.   SB3   . RL         .



 13.    (ViZDoom, ML-Agents, Gym)



Gymnasium   API . ViZDoom   FPS-. ML-Agents  Unity, curriculum, imitation, self-play. StarCraft (pysc2)  /,  . : SB3, CleanRL, RLlib, Sample Factory. : Gym ? Atari ? ViZDoom ? ML-Agents ?  .



 14. -  FPS, MMO  



FPS: YOLO ?  ? lead ?   ? ;  latency. MMO: UI, SM/BT, , , recovery; perception = CV + OCR; high-level  LLM. MOBA   . RTS  API  vision. Sandbox    LLM. :  ,  , recovery, ,  .



 15. -   



:  (BE, EAC, Vanguard),  (, ML), .  , , DLL, ,  ,  , injected input. Vision     memory-,   behavioral.    offline/research.



 16. Vision-Language Models  LLM-  



VLM:  +  ?  (, , ).   ,  . :  9598 %, LLM    . , , ReAct. SIMA  generalist  . Ƹ  ,  , , YOLO  .



 17. -:    



: OpenCV fishing; YOLO-aim; MAA-; Mineflayer (API-first); Peek-a-bot (RL in UE5); SIMA;  LLM-RPG. :   ;   ;  ;    .



 18. ,   



: Python; OpenCV, ultralytics, ONNX/TensorRT, EasyOCR; mss/dxcam, pydirectinput; PyTorch, Gymnasium, SB3; Roboflow, CVAT. : mss+OpenCV+pydirectinput+ultralytics.    TensorRT, , , .



 19.  -  



: generalist-; agentic AI  ; sample efficiency;   ;  ;   . :  perception, , , .



 20.    



: vision ;   end-to-end;     ;  .  : OpenCV ? SM ? YOLO ? Gym/PPO ? LLM-.    -; ;  offline.         .



 A.     



A.1.    



while running:

frame = capture.get_frame()

if frame is None:

continue

state = vision.analyze(frame)

actions = decisions.decide(state)

input_ctrl.execute(actions)

logger.tick(state, actions)

if emergency_stop_pressed():

break



A.2.    



bot/

main.py

capture.py

vision/

detector.py

ocr.py

state.py

decisions/

state_machine.py

policies.py

input_controller.py

config.yaml

templates/

weights/

logs/



A.3. -     



 Dry-run vision  

     

   

      (windowed/borderless)

  

        



 B.   



Agent, Behavioral cloning, Bounding box, Curriculum learning, DPI awareness,

End-to-end, Gymnasium, IoU, Latency, mAP, NMS, OCR, Partial observability,

Policy, Reward shaping, ROI, State machine, Template matching, VLM, YOLO,

PPO, DQN, TensorRT, ONNX, ByteTrack, Kalman filter, ReAct, SIMA.



 C.  



: OpenCV, ultralytics YOLO, Gymnasium, Stable-Baselines3, ViZDoom,

Unity ML-Agents, mss, dxcam.



: Sutton & Barto  Reinforcement Learning: An Introduction.



: DQN, AlphaGo, AlphaStar, OpenAI Five, SIMA (DeepMind).



:          .



 



      .

  .    

   -    .



    ˨ 



 -----------------------------------------------------------

:  state machine

 -----------------------------------------------------------



          .         : , ,  , , , , , ,  ,   , .



      (,   perception).        ,    ,     state object.        timestamp   .     hierarchical state machines  behaviour trees.



 -----------------------------------------------------------

:      YOLO

 -----------------------------------------------------------



   80 %  .     ,     ,   .    ,  ,  .  :     ,    .  active learning:              .  .      ,    .



 -----------------------------------------------------------

:  latency 

 -----------------------------------------------------------



          .   : , copy  GPU, , ,  ,  .      ,          .  FP16/INT8, TensorRT,  input size, .       ,     .



 -----------------------------------------------------------

: reward shaping  reward hacking

 -----------------------------------------------------------



   ,     .    ,        .        .         ( , ,  ). Curriculum        .



 -----------------------------------------------------------

:  LLM +  CV

 -----------------------------------------------------------



: perception    . LLM    ( , OCR-,  SM-,  )   .    JSON    .     . LLM   : ,  ,  UI,  .   ,   .



 -----------------------------------------------------------

 : vision-aim  FPS

 -----------------------------------------------------------



 :    latency ? YOLO (nano/small + TensorRT) ? NMS ?    ?    (Kalman  -) ?          ?        raw input ?     ?        ?     .    . :   ,  ,  ,    .        -.



 -----------------------------------------------------------

 : - MMO  vision

 -----------------------------------------------------------



:   /;  (,  , ); ;   ;  ;    ; -AFK (  ). Perception:  /NPC, /template  , OCR  HP   . SM  behaviour tree  .         UI    .



 -----------------------------------------------------------

 :  PPO  ViZDoom

 -----------------------------------------------------------



   Gymnasium,  observation ( vs game variables), frame skip, reward shaping   (, ). Stable-Baselines3 PPO: n_steps, batch_size, learning_rate, ent_coef, gamma.   reward  episode length.   curriculum-.           ;     shaping.



  



:   ,      Python?

:  OpenCV  mss.  ,   FPS .  template matching     single-player     pydirectinput.   .



:   ?

:   OpenCV  .  YOLO     NVIDIA  CUDA/TensorRT.    GPU   .



:      ?

: ,    UI,    .          .



:    ?

:     -    ToS-.   single-player, offline,   research-.



:  LLM  state machine?

:        /.  ,   .  .



:     YOLO   ?

:   3001500        ,   .     .



:   ?

:  ultralytics  Gymnasium; Sutton & Barto;    DeepMind   ;   open-source .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.

 Confidence YOLO:   0.350.5  .

 IoU NMS: 0.450.6  .

   :   ~0.150.2      .

  ROI:  ;  ROI     false positives.

          .

      .

        .

        ,    .

  profile  : , ROI, hotkeys, sens, .



:         monitor index       .

:   OCR   ROI      ,   .

:  imitation learning      ,     ,     .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.

 Confidence YOLO:   0.350.5  .

 IoU NMS: 0.450.6  .

   :   ~0.150.2      .

  ROI:  ;  ROI     false positives.

          .

      .

        .

        ,    .

  profile  : , ROI, hotkeys, sens, .



:         monitor index       .

:   OCR   ROI      ,   .

:  imitation learning      ,     ,     .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.

 Confidence YOLO:   0.350.5  .

 IoU NMS: 0.450.6  .

   :   ~0.150.2      .

  ROI:  ;  ROI     false positives.

          .

      .

        .

        ,    .

  profile  : , ROI, hotkeys, sens, .



:         monitor index       .

:   OCR   ROI      ,   .

:  imitation learning      ,     ,     .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.

 Confidence YOLO:   0.350.5  .

 IoU NMS: 0.450.6  .

   :   ~0.150.2      .

  ROI:  ;  ROI     false positives.

          .

      .

        .

        ,    .

  profile  : , ROI, hotkeys, sens, .



:         monitor index       .

:   OCR   ROI      ,   .

:  imitation learning      ,     ,     .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.

 Confidence YOLO:   0.350.5  .

 IoU NMS: 0.450.6  .

   :   ~0.150.2      .

  ROI:  ;  ROI     false positives.

          .

      .

        .

        ,    .

  profile  : , ROI, hotkeys, sens, .



:         monitor index       .

:   OCR   ROI      ,   .

:  imitation learning      ,     ,     .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.

 Confidence YOLO:   0.350.5  .

 IoU NMS: 0.450.6  .

   :   ~0.150.2      .

  ROI:  ;  ROI     false positives.

          .

      .

        .

        ,    .

  profile  : , ROI, hotkeys, sens, .



:         monitor index       .

:   OCR   ROI      ,   .

:  imitation learning      ,     ,     .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.

 Confidence YOLO:   0.350.5  .

 IoU NMS: 0.450.6  .

   :   ~0.150.2      .

  ROI:  ;  ROI     false positives.

          .

      .

        .

        ,    .

  profile  : , ROI, hotkeys, sens, .



:         monitor index       .

:   OCR   ROI      ,   .

:  imitation learning      ,     ,     .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.

 Confidence YOLO:   0.350.5  .

 IoU NMS: 0.450.6  .

   :   ~0.150.2      .

  ROI:  ;  ROI     false positives.

          .

      .

        .

        ,    .

  profile  : , ROI, hotkeys, sens, .



:         monitor index       .

:   OCR   ROI      ,   .

:  imitation learning      ,     ,     .



 -------------------------------------------------------------------------------

 :    

 -------------------------------------------------------------------------------

  FPS perception   : 1020;  FPS: 3060+.




  .


   .

   ,     (https://www.litres.ru/book/uchitel-nachalnoy/programmy-dlia-prokhozhdeniia-komp-iuternykh-igr-boty-ii-74457794/)  .

      Visa, MasterCard, Maestro,    ,   ,     ,  PayPal, WebMoney, ., QIWI ,       .


