anuashok/ocr-captcha-v2

1---2tags:3- vision4- ocr5- trocr6- pytorch7license: apache-2.08datasets:9- custom-captcha-dataset10metrics:11- cer12model_name: anuashok/ocr-captcha-v213base_model:14- microsoft/trocr-base-printed15---16 17# anuashok/ocr-captcha-v218 19This model is a fine-tuned version of [microsoft/trocr-base-printed](https://huggingface.co/microsoft/trocr-base-printed) on your custom dataset.20captchas like21 22 23![image/png](https://cdn-uploads.huggingface.co/production/uploads/6569b4be1bac1166939f86b2/urZTYpc7f5ZkC5qhUf_5l.png)24 25## Training Summary26 27- **CER (Character Error Rate)**: 0.0202593192868719628- **Hyperparameters**:29  - **Learning Rate**: 1.1081459294764632e-0530  - **Batch Size**: 431  - **Num Epochs**: 332  - **Warmup Ratio**: 0.0786313477415362833  - **Weight Decay**: 0.0624815282502137334  - **Num Beams**: 635  - **Length Penalty**: 0.509510072517366236 37## Usage38 39```python40from transformers import VisionEncoderDecoderModel, TrOCRProcessor41import torch42from PIL import Image43 44# Load model and processor45processor = TrOCRProcessor.from_pretrained("anuashok/ocr-captcha-v2")46model = VisionEncoderDecoderModel.from_pretrained("anuashok/ocr-captcha-v2")47 48# Load image49image = Image.open('path_to_your_image.jpg').convert("RGB")50 51# Prepare image52pixel_values = processor(image, return_tensors="pt").pixel_values53 54# Generate text55generated_ids = model.generate(pixel_values)56generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]57print(generated_text)