Skip to content

Latest commit

 

History

History
74 lines (51 loc) · 2.48 KB

README.md

File metadata and controls

74 lines (51 loc) · 2.48 KB
Raw

Metin2 (MMORPG) Vision Bot

Automatic Bot for Massive Dungeons Passing based on Windows API, YoloV8 object detection, statistical methods from OpenCV, Tesseract-OCR and spaCy virtualised with Hyper-V (Win11+CUDA)

Demo

demo_dung_stage4_5.mp4
demo_idle_metins.mp4
demo_dung_metins.mp4

Key Packages

  • Keyboard and Mouse controll: pynput
  • Custom Data Annotation: Roboflow
  • Object detection: ultralytics, OpenCV
  • Messages handling: Tesseract-OCR, spaCy

Funcionalities

  • Autonomic Dungeon Passing
  • Autonomic Metin Stones Destroying among all of the channels
  • Idle Exp

All of those modes can handle boundary situations such as game crash or logging out.

YoloV8

Model prepared with Ultralytics YOLOv8.0.196 🚀 Python-3.10.12 torch-2.1.0+cu121 CUDA:0 (Tesla T4, 15102MiB)


Class Images Instances Precision* Recall* mAP50* mAP50-95*
all 31 36 0.993 0.939 0.974 0.828
boss_gnoll_cpt 31 11 0.983 0.818 0.931 0.739
metin_polany 31 12 0.996 1 0.995 0.867
npc_straznik 31 13 0.999 1 0.995 0.877

* - Box level


Custom Dataset

Roboflow Dataset Overview (316 images, 800x600)

Augmentation

  • Outputs per training example: 3
  • Flip: Horizontal
  • Blur: Up to 1px

Annotation Heatmap

Direction Recognition

Grayscale Threshold, Contours Recognition and not triangular shape filter