Spaces:
Sleeping
Sleeping
Add building an image captioning app
Browse files
app.py
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import io
|
| 3 |
+
import IPython.display
|
| 4 |
+
from PIL import Image
|
| 5 |
+
import base64
|
| 6 |
+
|
| 7 |
+
from transformers import pipeline
|
| 8 |
+
import gradio as gr
|
| 9 |
+
|
| 10 |
+
hf_api_key = os.environ['HF_API_KEY']
|
| 11 |
+
|
| 12 |
+
get_completion = pipeline("ner", model="Salesforce/blip-image-captioning-base")
|
| 13 |
+
|
| 14 |
+
def image_to_base64_str(pil_image):
|
| 15 |
+
byte_arr = io.BytesIO()
|
| 16 |
+
pil_image.save(byte_arr, format='PNG')
|
| 17 |
+
byte_arr = byte_arr.getvalue()
|
| 18 |
+
|
| 19 |
+
return str(base64.b64encode(byte_arr).decode('utf-8'))
|
| 20 |
+
|
| 21 |
+
def captioner(image):
|
| 22 |
+
base64_image = image_to_base64_str(image)
|
| 23 |
+
result = get_completion(base64_image)
|
| 24 |
+
|
| 25 |
+
return result[0]['generated_text']
|
| 26 |
+
|
| 27 |
+
gr.close_all()
|
| 28 |
+
demo = gr.Interface(fn=captioner,
|
| 29 |
+
inputs=[gr.Image(label="Upload image", type="pil")],
|
| 30 |
+
outputs=[gr.Textbox(label="Caption")],
|
| 31 |
+
title="Image Captioning with BLIP",
|
| 32 |
+
description="Caption any image using the BLIP model",
|
| 33 |
+
allow_flagging="never",
|
| 34 |
+
examples=["christmas_dog.jpeg", "bird_flight.jpeg", "cow.jpeg"])
|
| 35 |
+
|
| 36 |
+
demo.launch(
|
| 37 |
+
share=True,
|
| 38 |
+
#server_port=int(os.environ['PORT1'])
|
| 39 |
+
)
|