maharleeeka / certificates

object detection on certificates

1. Set up YOLO with this repository -> https://github.com/AlexeyAB/darknet
2. Put objectDetector.py and ocr.py inside x64 folder
3. Make a folder named validation and put sequenceMatcher.py inside

about each file

objectDetector.py

• for object detection
• asks used to input the path of the image to test
• you can change the command value depending on your need
  • darknet_no_gput.exe for windows with no gpu
  • darknet.exe for windows with gpu

ocr.py

• for text recognition
• asks user to input the path of the image to test (the same path you used for object detection)
• change PATH and IMAGES_PATH values
  • PATH = whole path for the x64 folder
  • IMAGES_PATH = you can make another folder inside x64 to put all preprocessed images and final results`
  • line 64 is where you save the final result, please change the path in accordance to what you have
• if you want to save denoised and grayscaled images, uncomment lines no. 17 and 21

sequenceMatcher.py

• this code was used to test the 30 text recognition results (results are in text file for each image) in bulk
• sample test filenames are TEST_1.txt, TEST_2.txt, TEST_30.txt`
• depending on how many text files you want to test, you can change the
• i value in line 7 = which file to start
• while (i <= __): in line 9 = in the blank, change the number to until which file you want to stop testing
• inside a test file are the class name (in number), original text, and actual result separated by the character |, a newline for each class

---

• yolo-obj.cfg and data folder were used for training the detector
• YOLO weights are available in https://bit.ly/2XA0fnW