Measurement for OCR Prediction
Budget: €30 – €250 EUR
In this project we are Measuring quality of Image for OCR. We are testing of OCR performance against image degradations, to identify the most significant contributors to accuracy reduction. So if the input image quality is bad OCR won't be accurate. We are given multiple test images that have different level of accuracy and have different factors effecting the image like for example noise, blurriness, shadow, entropy, warp, etc.
*** In this project we aren't measuring OCR *** , We have to come up with some sort of measure that will give a prediction weather if OCR will work independent to ocr. We are going to do the OCR and compare the actual ocr accuracy to the prediction we came up with. OCR is not part of the quality measure, It should stand independent
We need to different detection like contrast score detection, blurness detection, entropy score detection plus warpness detection through text before giving to OCR. After that we can set thresholds for each detection to determine if its good for ocr or not.
We are testing of OCR performance against image degradations, to identify the most significant contributors to accuracy reduction.
We aren't changing anything on the image, We are just measuring different factors that would effect the image quality for OCR and predicting the OCR
*** In this project we aren't measuring OCR *** , We have to come up with some sort of measure that will give a prediction weather if OCR will work independent to ocr. We are going to do the OCR and compare the actual ocr accuracy to the prediction we came up with. OCR is not part of the quality measure, It should stand independent
We need to different detection like contrast score detection, blurness detection, entropy score detection plus warpness detection through text before giving to OCR. After that we can set thresholds for each detection to determine if its good for ocr or not.
We are testing of OCR performance against image degradations, to identify the most significant contributors to accuracy reduction.
We aren't changing anything on the image, We are just measuring different factors that would effect the image quality for OCR and predicting the OCR