anuashok/ocr captcha v3 This model is a fine tuned version of microsoft/trocr base printed on Captchas of the type shown below Training Summary CER (Character Error Rate) : 0.01394585726004922 Hyperparameters : Learning Rate : 1.5078922700531405e 05 Batch Size : 16 Num Epochs : 7 Warmup Ratio : 0.14813004670666596 Weight Decay : 0.017176551931326833 Num Beams : 2 Length Penalty : 1.3612823161368288 Usage python from transformers import VisionEncoderDecoderModel, TrOCRProcessor import torch from PIL import Image Load model and processor processor = TrOCRProcessor.from pretrained("anuashok/ocr captcha v3") model = VisionEncoderDecoderModel.from pretrained("anuashok/ocr captcha v3") Load image image = Image.open('path to your image.jpg').convert("RGB") Load and preprocess image for display image = Image.open(image path).convert("RGBA") Create white background background = Image.new("RGBA", image.size, (255, 255, 255)) combined = Image.alpha composite(background, image).convert("RGB") Prepare image pixel values = processor(combined, return tensors="pt").pixel values Generate text generated ids = model.generate(pixel values) generated text = processor.batch decode(generated ids, skip sp…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy