Myenny / d2vs

a library to assist with the automation of d2 using only pixels, keyboard, and mouse

what is this

diablo 2 vision system is a library to assist with the automation of d2, using only pixels, keyboard and mouse

installation

$ pip install d2vs

NOTE: this can run via CPU, but via GPU is far superior. You must install CUDA and the appropriate python jazz to get that working, for me with CUDA10.1:

$ conda install torch torchvision cudatoolkit=10.1 -c pytorch

usage

(586_gold.png)

import numpy as np
from d2vs.ocr import OCR
from PIL import Image

# Initiate OCR
ocr = OCR()

# Load an Image
img = Image.open("586_gold.png")

# Scan the image
bounds, text, item_type = ocr.read(img)

# Print out the data for demo purposes
top_left, top_right, bottom_right, bottom_left = bounds
print(top_left, top_right, bottom_right, bottom_left)
# ([2, 2], [158, 2], [158, 32], [2, 32])

print(text)
# '586 Gold'

print(item_type)
# 'Normal'

project goals

• Have fun automating single player! Not for profit
• OCR with near 100% accuracy
• Visually determine where you are in game, area level and world coordinate system
• Click from world coords to screen coords
• Path through unexplored areas to a goal
• Facilitate complete d2 bot from lvl 1 to 99
• Pick it

an example map reading from d2vs of black marsh

nodes for the static map Harrogath

development

setup

$ git clone ...
$ pip install -r requirements.dev.txt

running tests

$ pytest