Delete Readme.md
Browse files
Readme.md
DELETED
|
@@ -1,37 +0,0 @@
|
|
| 1 |
-
## Usage method:
|
| 2 |
-
```python
|
| 3 |
-
from transformers import VisionEncoderDecoderModel, ViTImageProcessor, AutoTokenizer
|
| 4 |
-
import torch
|
| 5 |
-
from PIL import Image
|
| 6 |
-
|
| 7 |
-
model = VisionEncoderDecoderModel.from_pretrained("AIris-Channel/vit-gpt2-verifycode-caption")
|
| 8 |
-
feature_extractor = ViTImageProcessor.from_pretrained("AIris-Channel/vit-gpt2-verifycode-caption")
|
| 9 |
-
tokenizer = AutoTokenizer.from_pretrained("AIris-Channel/vit-gpt2-verifycode-caption")
|
| 10 |
-
|
| 11 |
-
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 12 |
-
model.to(device)
|
| 13 |
-
|
| 14 |
-
max_length = 16
|
| 15 |
-
num_beams = 4
|
| 16 |
-
gen_kwargs = {"max_length": max_length, "num_beams": num_beams}
|
| 17 |
-
def predict_step(image_paths):
|
| 18 |
-
images = []
|
| 19 |
-
for image_path in image_paths:
|
| 20 |
-
i_image = Image.open(image_path)
|
| 21 |
-
if i_image.mode != "RGB":
|
| 22 |
-
i_image = i_image.convert(mode="RGB")
|
| 23 |
-
|
| 24 |
-
images.append(i_image)
|
| 25 |
-
|
| 26 |
-
pixel_values = feature_extractor(images=images, return_tensors="pt").pixel_values
|
| 27 |
-
pixel_values = pixel_values.to(device)
|
| 28 |
-
|
| 29 |
-
output_ids = model.generate(pixel_values, **gen_kwargs)
|
| 30 |
-
|
| 31 |
-
preds = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
|
| 32 |
-
preds = [pred.strip() for pred in preds]
|
| 33 |
-
return preds
|
| 34 |
-
|
| 35 |
-
pred=predict_step(['ZZZTVESE.jpg'])
|
| 36 |
-
print(pred) #zzztvese
|
| 37 |
-
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|