Once you get a result back from the server, you can click the dictionary button to look up the word in the server's dictionary. Once you get a result, you can click on each character to see the next two candidates returned from the server, in case there was a near miss. OCR-JPN works best on Gothic-type fonts (bold and blocky instead of script-like or calligraphic). The script will continue to search until it finds a "good enough" score, at which point it will compare the input image to characters that have been deemed similar to the "good enough" character.
The program then performs an analysis to count the number of unique negative spaces present in the thresholded image and uses that as a starting point in a database of around 3000 Japanese characters. The server then crops the image and performs several manipulations using PIL (Python Imaging Library) that resize, threshold, and crop the image to prepare it for comparison. Please note that no images from your session are saved to the server. This will send a screenshot of your tab to the server. Draw the window over the character(s) you wish to recognize, trying to get a clean crop (although it is forgiving if you catch a little extra of the characters next to, but not above or below, your selection). Once you click on the extension icon, a dialog window will open on the current page. The goal of OCR-JPN is to turn Japanese text on images into actual text so that you can look it up in a dictionary or copy/paste it into a flashcard program or notes. OCR stands for Optical Character Recognition and is used to turn pictures containing representations of text into actual text that can be manipulated like text and not an image. One way to solve this problem is to use OCR.
The fantastic Chrome extension rikaikun lets you hover over Japanese text and get a dictionary popup. OCR-JPN was built by Siena Aguayo as her final project at Hackbright Academy during the fall of 2013. You can then copy/paste the text into your favorite dictionary, or perform a lookup on the spot using a built-in implementation of JMdict. OCR-JPN is a Chrome extension that lets you recognize Japanese characters in images you find around the web.
0 kommentar(er)
