File size: 1,086 Bytes
cfd8b54 bbfa147 f42a674 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 |
---
tags:
- mgp-str
- text-recognition
---
This is an ONNX export of the [MGP-STR](https://huggingface.co/docs/transformers/model_doc/mgp-str) model for text recognition.
It can be run as follows:
```
import onnxruntime as ort
from transformers import MgpstrProcessor
processor = MgpstrProcessor.from_pretrained("alibaba-damo/mgp-str-base")
mpg_str_onnx = hf_hub_download(repo_id="ml6team/mgp-str-onnx", filename="mgp-str.onnx", repo_type="model")
providers = [('CUDAExecutionProvider', {"cudnn_conv_algo_search": "DEFAULT"}), 'CPUExecutionProvider'] if ort.get_device() == 'GPU' else ['CPUExecutionProvider']
session = ort.InferenceSession(
mpg_str_onnx, providers=providers,
)
image = Image.open("path_to_your_image")convert("RGB")
pixel_values = processor(images=image, return_tensors="pt").pixel_values
inputs = {"pixel_values": pixel_values.numpy()}
warmup = mgp_str_session.run(None, inputs)
outputs = session.run(None, inputs)
outputs = [torch.tensor(i) for i in outputs]
out_strs = processor.batch_decode(tuple(outputs))
print(out_strs["generated_text"])
``` |