sneed

app.py

@@ -1,27 +1,31 @@
 import gradio as gr
 from transformers import TrOCRProcessor, VisionEncoderDecoderModel
-import requests
+import requests, re, base64
 from PIL import Image
-
+from io import BytesIO
 processor = TrOCRProcessor.from_pretrained("microsoft/trocr-small-printed")
 model = VisionEncoderDecoderModel.from_pretrained("pENrknSoysneed/8kun-captcha-ocr-meme")
 
-# load image examples
-urls = [
-    'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/nfcb5.png',
-    'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/p57fn.png',
-    'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/w2yp7.png',
-    'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/pme86.png',
-    'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/w4nfx.png',
-    'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/nf8b8.png'
-]
-for idx, url in enumerate(urls):
-  image = Image.open(requests.get(url, stream=True).raw)
-  image.save(f"image_{idx}.png")
+# # load image examples
+# urls = [
+#     'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/nfcb5.png',
+#     'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/p57fn.png',
+#     'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/w2yp7.png',
+#     'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/pme86.png',
+#     'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/w4nfx.png',
+#     'https://storage.googleapis.com/trocr-captcha.appspot.com/captcha_images_v2/nf8b8.png'
+# ]
+# for idx, url in enumerate(urls):
+#   image = Image.open(requests.get(url, stream=True).raw)
+#   image.save(f"image_{idx}.png")
 
 def process_image(image):
     # prepare image
-    pixel_values = processor(image, return_tensors="pt").pixel_values
+    image_data = re.sub('^data:image/.+;base64,', '', image)
+    im = Image.open(BytesIO(base64.b64decode(image_data)))
+    
+    #Take's the picture
+    pixel_values = processor(im, return_tensors="pt").pixel_values
 
     # generate (no beam search)
     generated_ids = model.generate(pixel_values)
@@ -39,8 +43,10 @@ description = "Due to events. in 8chan staff moderation. I am attacking it. The
 #css = """.output_image, .input_image {height: 600px !important}"""
 
 iface = gr.Interface(fn=process_image, 
-                     inputs=gr.inputs.Image(type="pil"), 
-                     outputs=gr.outputs.Textbox(),
+                    #  inputs=gr.inputs.Image(type="pil"), 
+                    inputs=gr.Textbox(placeholder="base64 string (right-click => copy-link) ...")
+                    outputs=gr.outputs.Textbox(),
+                    
                      title=title,
                      description=description,
                     #  article=article,