|
import transformers |
|
import torch |
|
import gradio as gr |
|
import requests |
|
|
|
from transformers import BlipForImageTextRetrieval |
|
from transformers import AutoProcessor |
|
from transformers.utils import logging |
|
from PIL import Image |
|
logging.set_verbosity_error() |
|
|
|
model = BlipForImageTextRetrieval.from_pretrained("Salesforce/blip-itm-base-coco") |
|
processor = AutoProcessor.from_pretrained("Salesforce/blip-itm-base-coco") |
|
|
|
def process_image(input_type, image_url, image_upload, text): |
|
if input_type == "URL": |
|
raw_image = Image.open(requests.get(image_url, stream=True).raw).convert('RGB') |
|
else: |
|
raw_image = image_upload |
|
|
|
inputs = processor(images=raw_image, text=text, return_tensors="pt") |
|
itm_scores = model(**inputs)[0] |
|
itm_score = torch.nn.functional.softmax(itm_scores,dim=1) |
|
itm_score = itm_score[0][1] |
|
print(itm_score) |
|
|
|
if itm_score <=.35: |
|
cmnt = "which is not that great. Try again." |
|
elif itm_score <= .75: |
|
cmnt = "which is good. But you can improve it. Try again." |
|
elif itm_score == 1.0: |
|
cmnt = "and that is an unbelievable perfect score. You have achieved the near impossible. Congratulations" |
|
else: |
|
cmnt = "which is excellent. Can you improve on it?" |
|
|
|
formatted_text = ( |
|
f"""<div><h1 style='text-align: center; font-size: 80px; color: blue;'> |
|
Your decription score is <span style='font-size: 100px; color: orange;'>{itm_score*100:.2f}/100</span> {cmnt} |
|
</h1></div>""" |
|
) |
|
return formatted_text |
|
|
|
def display_image_from_url(image_url): |
|
if image_url: |
|
image = Image.open(requests.get(image_url, stream=True).raw).convert('RGB') |
|
return image |
|
return None |
|
|
|
def toggle_inputs(input_type): |
|
if input_type == "URL": |
|
return gr.update(visible=True), gr.update(visible=True), gr.update(visible=False), gr.update(visible=True) |
|
else: |
|
return gr.update(visible=False), gr.update(visible=False), gr.update(visible=True), gr.update(visible=True) |
|
|
|
with gr.Blocks() as demo: |
|
gr.Markdown( |
|
""" |
|
# Challenge yourself by describing the image - test & demo app by Srinivas.V.. |
|
Paste either URL of an image or upload the image, describe the image best and submit to know your score. |
|
""") |
|
|
|
input_type = gr.Radio(choices=["URL", "Upload"], label="Input Type") |
|
image_url = gr.Textbox(label="Image URL", visible=False) |
|
url_image = gr.Image(type="pil", label="URL Image", visible=False) |
|
image_upload = gr.Image(type="pil", label="Upload Image", visible=False) |
|
description = gr.Textbox(label="Describe the image", visible=False, lines=3) |
|
|
|
input_type.change(fn=toggle_inputs, inputs=input_type, outputs=[image_url, url_image, image_upload, description]) |
|
image_url.change(fn=display_image_from_url, inputs=image_url, outputs=url_image) |
|
|
|
submit_btn = gr.Button("Submit") |
|
processed_image = gr.HTML(label="Your challenge result") |
|
submit_btn.click(fn=process_image, inputs=[input_type, image_url, image_upload, description], outputs=processed_image) |
|
|
|
demo.launch(debug=True, share=True) |