Upload 4 files

app.py

@@ -0,0 +1,51 @@
+from transformers import TrOCRProcessor, VisionEncoderDecoderModel
+from PIL import Image
+import requests
+import gradio as gr 
+
+# app title
+title = "ScriptSense"
+description = "<p style='text-align: center; font-size: 22px; font-weight: bold;'>design and crafted by aryan verma</p>"
+article = "<p style='text-align: center; font-size: 14px;'>aryan verma | 241306064</p>"
+css = "footer {display: none !important;}"
+
+# sample images
+examples = [
+    ["", None, "images/1.jpg"],
+    ["", None, "images/sample-handwritten-2.PNG"]
+]
+
+#you can load any model from huggingface
+processor = TrOCRProcessor.from_pretrained('microsoft/trocr-large-handwritten')
+model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-large-handwritten')
+
+# prediction function for handwritting
+def predict(ImageUrl,imgDraw,imgUplod):
+
+  #fetch the image from url or handwritten canvas or the uplaoded image
+  if ImageUrl :
+    image = Image.open(requests.get(ImageUrl, stream=True).raw).convert("RGB")
+  elif imgDraw : 
+    image = imgDraw.convert("RGB")
+  else :
+    image = imgUplod.convert("RGB")
+
+  #predict the image using microsoft/trocr-large-handwritten model loaded earlier
+  pixel_values = processor(images=image, return_tensors="pt").pixel_values
+  generated_ids = model.generate(pixel_values)
+  generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+  return generated_text
+
+#gradio interface
+interface = gr.Interface(
+    fn=predict, 
+    inputs=["text",gr.Sketchpad(type="pil",shape=(500, 500)),gr.Image(type="pil")], 
+    outputs="text", 
+    title=title,
+    description=description,
+    article=article,
+    examples=examples,
+    css=css
+)
+interface.launch()
+

requirements.txt

@@ -0,0 +1,9 @@
+transformers
+torch
+gradio==3.50.2
+pillow
+requests
+fastapi==0.104.1
+starlette==0.27.0
+pydantic==1.10.13
+jinja2<3.1.4