https://github.com/Madoshakalaka/whichimg

blazing fast template matching when possible images are all known
https://github.com/Madoshakalaka/whichimg

Last synced: 11 days ago
JSON representation

blazing fast template matching when possible images are all known

• Host: GitHub
• URL: https://github.com/Madoshakalaka/whichimg
• Owner: Madoshakalaka
• License: mit
• Created: 2019-07-01T07:54:12.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2019-07-09T16:05:33.000Z (over 5 years ago)
• Last Synced: 2024-05-01T14:38:04.984Z (7 months ago)
• Language: Python
• Size: 47.9 KB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-blazingly-fast - whichimg - blazing fast template matching when possible images are all known (Python)

README

        
# whichimg

![travis-badge](https://travis-ci.org/Madoshakalaka/whichimg.svg?branch=master)

blazing fast template matching when possible images are all known. Handy tool for GUI scripting.

## How to Use

It supports python 3.6 +

`pip install whichimg`

```python

from whichimg import ImageTeller

teller = ImageTeller([img1, img2, img3, img4]) # numpy arrays

teller.tell(secret_img) # returns 0 or 1 or 2 or 3 or -1 for not found

```

if you're sure `secret_img` is one of the images, you can set keyword argument `surprises` to `False`. This will give you a minor performance gain.

```python

teller = ImageTeller([img1, img2, img3, img4], surprises=False)

```

This is equivalent to the following naive approach

```python

def naive_tell(images, sample_img):

    for i, img in enumerate(images):

        if np.array_equal(img, sample_img):

            return i

    return -1

naive_tell([img1, img2, img3, img4], secret_img)

```

## Note

It's **not** generally faster than the naive approach. I thought my approach was faster and spent a week writing this shit though. lmfao.

They're about equally fast on the `tests/fixtures` testing data I came up with (10x10 images). Through my rough testing, there could be a magnitude of performance gain when there are many possible images (>10) and when the images are larger (200x200 for example).