https://github.com/developer239/simple-color-batch-cropper

Automatically crops positive/negative samples for your haar cascades. Example uses images from Swords and Souls web game. You can also use this library to mark your images.
https://github.com/developer239/simple-color-batch-cropper

Last synced: about 1 year ago
JSON representation

Automatically crops positive/negative samples for your haar cascades. Example uses images from Swords and Souls web game. You can also use this library to mark your images.

• Host: GitHub
• URL: https://github.com/developer239/simple-color-batch-cropper
• Owner: developer239
• Created: 2018-01-06T20:30:51.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2018-02-04T17:47:27.000Z (about 8 years ago)
• Last Synced: 2025-01-04T21:41:42.835Z (about 1 year ago)
• Language: JavaScript
• Homepage:
• Size: 2.78 MB
• Stars: 1
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: readme.MD

Awesome Lists containing this project

README

          
# Simple Color Batch cropper

### Installation

```

$ yarn install

```

Make sure to have OpenCV 3+ (extra modules are optional) installed on your system https://github.com/opencv/opencv/releases/.

More information can be found here: [opencv4nodejs](https://github.com/justadudewhohacks/opencv4nodejs#how-to-install)

### What does it do

#### Positive images

1) When you need to generate large amount of positive samples of all _shooting targets_ automatically

![1](https://github.com/developer239/simple-color-batch-cropper/blob/master/documentation/images/1.png?raw=true)

2) This script will generate binary image

![2](https://github.com/developer239/simple-color-batch-cropper/blob/master/documentation/images/2.png?raw=true)

3) Then it will find all shooting targets for you

![3](https://github.com/developer239/simple-color-batch-cropper/blob/master/documentation/images/3.png?raw=true)

4) Then it will draw bounding box around all targets

![4](https://github.com/developer239/simple-color-batch-cropper/blob/master/documentation/images/4.png?raw=true)

#### Negative images

`src/negative.js` will crop random locations from the original image so that none of the locations contain matches that you found using the `src/positive.js` file. This is useful when you want to generate large number of negative samples.

### How to use

* Put image that you want to process in `src/images` folder

* Create config variables in `src/config` folder. Every obect that you want to crop has its own config file. For example config for cropping apples: 

```

export const UPPER_COLOR = new Vec(200, 65, 255) // Upper color range !! IMPORTANT

export const LOWER_COLOR = new Vec(10, 10, 130) // Lower color range !! IMPORTANT

export const BLUR = 5 // Necessary for removing noise from image !! IMPORTANT

export const OFFSET = -50 // Top/left for positive image preview

export const REGION_SIZE = 100 // Positive image preview size

export const PATH = './src/images/withAppleSmall.jpg' // Path to image that will be parsed

```

* Import correct config in `src/positive.js` and `src/negative.js` files

* Run either `yarn positive` or `yarn negative` in order to see the result