Spaces:
Sleeping
Sleeping
Razzaqi3143
commited on
Commit
•
0b3d99b
1
Parent(s):
b50e4dc
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import VisionEncoderDecoderModel, ViTImageProcessor, AutoTokenizer
|
2 |
+
from PIL import Image
|
3 |
+
import torch
|
4 |
+
import gradio as gr
|
5 |
+
|
6 |
+
# Load your model and tokenizer
|
7 |
+
model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
|
8 |
+
processor = ViTImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
|
9 |
+
tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
|
10 |
+
|
11 |
+
# Function to generate captions from images
|
12 |
+
def generate_caption(image):
|
13 |
+
# Preprocess the image
|
14 |
+
pixel_values = processor(images=image, return_tensors="pt").pixel_values
|
15 |
+
|
16 |
+
# Generate captions
|
17 |
+
output_ids = model.generate(pixel_values, max_length=16, num_beams=4, return_dict_in_generate=True).sequences
|
18 |
+
caption = tokenizer.decode(output_ids[0], skip_special_tokens=True)
|
19 |
+
|
20 |
+
return caption
|
21 |
+
|
22 |
+
# Create a Gradio Interface
|
23 |
+
interface = gr.Interface(
|
24 |
+
fn=generate_caption,
|
25 |
+
inputs=gr.Image(type="pil"),
|
26 |
+
outputs=gr.Textbox(),
|
27 |
+
title="Image Caption Generator",
|
28 |
+
description="Upload an image and click 'Generate' to get a caption."
|
29 |
+
)
|
30 |
+
|
31 |
+
# Launch the app in Hugging Face Spaces
|
32 |
+
interface.launch()
|