|
|
import gradio as gr |
|
|
from transformers import BlipProcessor, BlipForConditionalGeneration, MarianMTModel, MarianTokenizer |
|
|
from PIL import Image |
|
|
import torch |
|
|
import random |
|
|
import datetime |
|
|
|
|
|
|
|
|
|
|
|
translator_model_ar = MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-en-ar") |
|
|
translator_tokenizer_ar = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-ar") |
|
|
|
|
|
|
|
|
model_path = "Projectt123/MuniVis" |
|
|
processor_en = BlipProcessor.from_pretrained(model_path) |
|
|
model_en = BlipForConditionalGeneration.from_pretrained(model_path) |
|
|
model_en.eval() |
|
|
|
|
|
|
|
|
@torch.no_grad() |
|
|
def describe_image(image, language): |
|
|
pil = image if isinstance(image, Image.Image) else Image.fromarray(image) |
|
|
|
|
|
|
|
|
inputs = processor_en(pil, return_tensors="pt") |
|
|
out_ids = model_en.generate(**inputs) |
|
|
en_desc = processor_en.decode(out_ids[0], skip_special_tokens=True) |
|
|
|
|
|
|
|
|
if str(language).lower().startswith(("ar", "arabic")) or language == "العربية": |
|
|
inputs_ar = translator_tokenizer_ar(en_desc, return_tensors="pt") |
|
|
ar_tokens = translator_model_ar.generate(**inputs_ar) |
|
|
ar_desc = translator_tokenizer_ar.decode(ar_tokens[0], skip_special_tokens=True) |
|
|
return ar_desc |
|
|
else: |
|
|
return en_desc |
|
|
|
|
|
ASIR_BBOX = { |
|
|
"lat_min": 17.0, "lat_max": 20.0, |
|
|
"lon_min": 41.5, "lon_max": 44.0, |
|
|
} |
|
|
|
|
|
def random_point_in_bbox(bbox): |
|
|
lat = random.uniform(bbox["lat_min"], bbox["lat_max"]) |
|
|
lon = random.uniform(bbox["lon_min"], bbox["lon_max"]) |
|
|
return round(lat, 6), round(lon, 6) |
|
|
|
|
|
def render_html(caption_text, lat, lon, language="English"): |
|
|
now = datetime.datetime.now() |
|
|
dt_str = now.strftime("%d %B %Y - %I:%M %p") |
|
|
|
|
|
is_ar = str(language).lower().startswith(("ar", "arabic")) or language == "العربية" |
|
|
dt_line = f"🕒 {'التاريخ والوقت' if is_ar else 'Date & Time'}: {dt_str}" |
|
|
loc_label = "📍 الموقع" if is_ar else "📍 Location" |
|
|
|
|
|
zoom = 16 |
|
|
osm_iframe = ( |
|
|
f'<iframe width="100%" height="300" frameborder="0" ' |
|
|
f'src="https://www.openstreetmap.org/export/embed.html?layer=mapnik&marker={lat}%2C{lon}&zoom={zoom}"></iframe>' |
|
|
f'<div style="font-size:12px;color:#aaa;margin-top:4px">{loc_label}: {lat}, {lon} ' |
|
|
f'• <a target="_blank" href="https://www.openstreetmap.org/?mlat={lat}&mlon={lon}#map={zoom}/{lat}/{lon}">Open map</a>' |
|
|
f'</div>' |
|
|
) |
|
|
|
|
|
return f""" |
|
|
<div style="font-family:system-ui,Segoe UI,Arial;color:#eee;line-height:1.5"> |
|
|
<div style="background:transparent;padding:14px 16px;border-radius:10px;margin-bottom:10px;"> |
|
|
<div style="font-size:16px;white-space:pre-wrap;">{caption_text}</div> |
|
|
</div> |
|
|
{osm_iframe} |
|
|
<div style="margin-top:8px;font-size:14px;">{dt_line}</div> |
|
|
</div> |
|
|
""" |
|
|
|
|
|
def ui_print_like(image, language): |
|
|
caption = describe_image(image, language) |
|
|
lat, lon = random_point_in_bbox(ASIR_BBOX) |
|
|
return render_html(caption, lat, lon, language) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
demo = gr.Interface( |
|
|
fn=ui_print_like, |
|
|
inputs=[ |
|
|
gr.Image(type="pil", label="Upload road image"), |
|
|
gr.Dropdown(choices=["Arabic", "English", "العربية"], value="English", label="Select Language") |
|
|
], |
|
|
outputs=gr.HTML(label="output"), |
|
|
title="MuniVis – Road Issue Detector (Asir)", |
|
|
description="Upload an image and get a description with a map (Asir random location) and date/time." |
|
|
) |
|
|
|
|
|
demo.launch(debug=True) |
|
|
|
|
|
|
|
|
|
