Merge pull request #2 from NicBonetto/streamlit-app

.gitattributes

@@ -0,0 +1,3 @@
+trocr-ocr/* filter=lfs diff=lfs merge=lfs -text
+trocr-ocr/** filter=lfs diff=lfs merge=lfs -text
+model/** filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,2 +1,13 @@
+---
+title: Light OCR Finetuned
+emoji: "📝"
+colorFrom: "purple"
+colorTo: "blue"
+sdk: streamlit
+sdk_version: "1.25.0"
+app_file: app.py
+pinned: true
+---
+
 # light-doc-ocr
 Lightweight Transformer-based OCR model for document text recognition, fine-tuned with PyTorch and Hugging Face.

app.py

@@ -0,0 +1,24 @@
+import streamlit
+from PIL import Image
+import torch
+from transformers import TrOCRProcessor, VisionEncoderDecoderModel
+
+MODEL_NAME = 'model/'
+processor = TrOCRProcessor.from_pretrained(MODEL_NAME)
+model = VisionEncoderDecoderModel.from_pretrained(MODEL_NAME)
+
+streamlit.title('Light OCR')
+
+uploaded_file = streamlit.file_uploader('Choose an image...', type=['png', 'jpg', 'jpeg'])
+if uploaded_file:
+    image = Image.open(uploaded_file).convert("RGB")
+    image = image.resize((384, 384))
+    streamlit.image(image, caption='Uploaded Image', use_column_width=True)
+    
+    pixel_values = processor(images=image, return_tensors='pt').pixel_values
+    output_ids = model.generate(pixel_values)
+    text = processor.batch_decode(output_ids, skip_special_tokens=True)[0]
+    
+    streamlit.subheader('Recognized Text')
+    streamlit.write(text)
+

model/config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:584e4b4106aa735f04ef9d5dd2e89250f3c98f36435a183fe98bd4888c4c5b2b
+size 1529

model/generation_config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66a55c9bff6d80e77c8deb6dba8dd79d867da689c7e0f1e1eddb265f8a92fb1b
+size 185

model/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d77161e6a5564a2d70e53b5dabfad12b67fb2e9bd7c3cc7555b1fe056bc8826
+size 246430696

model/preprocessor_config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5172006ffcaf0f407db91ac4ada30ad6ca86183fa37ce6059f966bfaffb880cb
+size 411

requirements.txt

@@ -5,4 +5,5 @@ datasets
 evaluate
 jiwer
 Pillow
-
+accelerator
+streamlit