gtraid: A Python repository from DraTeots

Utility to automatically parse Guardian Tales screenshots and build Excel spreadsheets with raid hits, utilizing computer vision (OpenCV) and text recognision (pyTesseract).

Currently the workflow assumed to be:

An admin has python and everything installed
He saves the raid screenshots to a folder
Then he runs this script, which recognizes all images and creates Excel file with results
The Excel file contains both images and recognized names/damages

Readiness:

name + damage image recognition
support multiple resolutions
auto mark possibly duplicated hit records
parse multiple raid screenshots
create output Excel file with results
calculate hit time, calculated from screenshot time
recognize boss (provide text name)
recognize team (provide text name of each member)

Future plans (enhanced workflow):

Have a discord bot
Admins uploads images to a discord channel
Images got processed and are automatically added to some DB.

This is possible only if some other developer will contribute

Installation

Install tesseract

https://github.com/UB-Mannheim/tesseract/wiki
Install python

https://anaconda.org/
Download this repo
(if you don't know how to use GitHub, you can download zipped file Code -> Download zip)
Open a python console in a folder with cloned/unzipped code and do:
```
pip install -m requirments.txt
```
Run test, to check you nailed it and script is working
```
 python test.py
```
You should see in details how a test image is processed (just press space)

Usage

In the directory with gt.py run

python gt.py <path to files>
python gt.py screenshot1.jpg screenshot2.jpg
python gt.py screenshot_folder/*

Flags:

-d, --debug - Enable debugging output. 0-none, 1-prints, 2-show images
-r, --report - Report folder (set blank for no report)
-o, --output - File name of resulting xlsx
-t, --tesseract - Full path to tesseract.exe

Results

You have a resulting file called by default result.xlsx

(!) Pink cells - means same damage with the same name happened before, probably duplicate reord

Problems

Something is not recognized

1 - Asian alphabet is known to cause problems.

2 - In GT UI not all the hits fit into hits window. So you have to choose to cut the first hit from the top or the last one from the bottom. Hits cut from the top may cause problem (see skewed up rectangles' placement):

3 - Something is not recognized sometimes. Just go over Excel spreadsheet and fix it manually

pip is not found

You have installed python without pip (python package manager). Either install anaconda (overkill, but reliable) or google how to install pip

Resolution is not found

This software needs coordinates how to crop images for each resolution. And there are myriads resolutions out here now. So probably this resolution is not yet supported. One needs to add cropping coordinates to dimensions.yaml file And if you successfully do this, don't forget to submit this changes to this repo.

Tesseract is not found

You need tesseract package installed to run. By default, the tesseract.exe location is assumed to be:

C:\Program Files\Tesseract-OCR\tesseract.exe

If it is installed in another location use -t flag.

While using -t flag don't forget about double backslash \\ for windows paths. Btw, slash / also works for windows paths and don't blow your mind... most of the times

Add your resolution

If the beast is not working because your resolution is unknown, you have to edit dimensions.yaml. The file contains information of how to cut and dissect a picture for each resolution. In particular, coordinates of cropping rectangles or "crop_rects" (as it named in the code)

All crop rectangles are given like:

hits_window:
   x_start: 431
   y_start: 148
   x_end: 1770
   y_end: 1003

Those are absolute coordinates of rectangles from left top corner of a picture. Just use one of the graphics editor which shows pixel coordinates under mouse (even paint should do it... I think)

hits_window Coordinates of a window with hits from your original screenshot. Should be this rectangle:

(!) black outer borders are important to recognize hits boxes

hit_image After each hit is cropped, the software dissects hit image like this:

name_rect - blue
party_rect - green
damage_rect - yellow
boss_rect - red

Debugging

When you run gt.py it creates "report" folder with all sub images it creates, you should see how it processes images and what it sees.

Or use --debug=2 flag. It will show you all interactively

python gt.py --debug=2 your_file.jpg

And report bugs and everything here!

DraTeots/gtraid