File size: 6,358 Bytes
a49af3b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 |
# -*- coding: utf-8 -*-
"""Untitled4.ipynb
Automatically generated by Colab.
Original file is located at
https://colab.research.google.com/drive/1LWcAEqZ865KCYhK-crQ3RKcFJEcH4xhD
"""
from huggingface_hub import login
login(token="hf_xxxx")
import torch
from transformers import Blip2Processor, Blip2ForConditionalGeneration
from PIL import Image
import gradio as gr
from io import BytesIO
import requests
# -----------------------------
# 1️⃣ Device and model setup
# -----------------------------
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")
# Use a smaller model for memory efficiency
model_name = "Salesforce/blip2-flan-t5-xl-coco" # Updated model with preprocessor_config.json
processor = Blip2Processor.from_pretrained(model_name)
model = Blip2ForConditionalGeneration.from_pretrained(model_name)
model.to(device)
# -----------------------------
# 2️⃣ Inference function
# -----------------------------
def analyze_eo_image(image, question):
"""
Analyze Earth Observation (EO) image with a natural language question.
"""
try:
if image is None or question.strip() == "":
return "Please upload an EO image and ask a question."
# Convert image to RGB if needed
if image.mode != "RGB":
image = image.convert("RGB")
# Prepare inputs
inputs = processor(image, question, return_tensors="pt").to(
device, torch.float16 if device == "cuda" else torch.float32
)
# Generate answer
out = model.generate(
**inputs,
max_new_tokens=80, # allow slightly longer explanations
do_sample=True,
temperature=0.7
)
# Decode and return
answer = processor.decode(out[0], skip_special_tokens=True)
return answer
except Exception as e:
return f"Error: {e}"
# -----------------------------
# 3️⃣ Optional: Analyze image from URL
# -----------------------------
def analyze_eo_url(url, question):
try:
response = requests.get(url)
image = Image.open(BytesIO(response.content)).convert("RGB")
return analyze_eo_image(image, question)
except Exception as e:
return f"Error loading image: {e}"
# -----------------------------
# 4️⃣ Gradio Interface
# -----------------------------
interface = gr.Interface(
fn=analyze_eo_image,
inputs=[
gr.Image(type="pil", label="Upload EO Image"),
gr.Textbox(label="Ask a Question about the EO Image")
],
outputs="text",
title="🌍 EO Image Analysis with Multimodal GPT-OSS",
description="""
Ask questions about Earth Observation (EO) images.
Powered by BLIP-2 + FLAN-T5 (small, memory-efficient).
Examples: "Identify land cover types", "Where is the river?",
"Has urban area expanded?".
""",
allow_flagging="never"
)
# Launch the interface (shareable link)
interface.launch(share=True)
import torch
from transformers import Blip2Processor, Blip2ForConditionalGeneration
from PIL import Image
import gradio as gr
from io import BytesIO
import requests
# -----------------------------
# Device and Model Setup
# -----------------------------
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")
model_name = "Salesforce/blip2-flan-t5-xl-coco" # public model
processor = Blip2Processor.from_pretrained(model_name)
model = Blip2ForConditionalGeneration.from_pretrained(model_name)
model.to(device)
# -----------------------------
# Inference Function
# -----------------------------
def analyze_eo_image(image, question):
if image is None or question.strip() == "":
return "Please upload an EO image and ask a question."
if image.mode != "RGB":
image = image.convert("RGB")
inputs = processor(image, question, return_tensors="pt").to(
device, torch.float16 if device=="cuda" else torch.float32
)
out = model.generate(
**inputs,
max_new_tokens=80,
do_sample=True,
temperature=0.7
)
return processor.decode(out[0], skip_special_tokens=True)
# -----------------------------
# Optional: URL input
# -----------------------------
def analyze_eo_url(url, question):
try:
response = requests.get(url)
image = Image.open(BytesIO(response.content)).convert("RGB")
return analyze_eo_image(image, question)
except Exception as e:
return f"Error loading image: {e}"
# -----------------------------
# Beautiful Gradio Layout
# -----------------------------
with gr.Blocks(title="🌍 EO Image Analysis") as demo:
gr.Markdown(
"""
# 🌍 Earth Observation Image Analysis
Ask questions about EO images using a multimodal AI model.
Powered by BLIP-2 + FLAN-T5.
**Examples:** "Identify land cover types", "Where is the river?", "Has urban area expanded?"
"""
)
with gr.Tabs():
with gr.Tab("Upload Image"):
with gr.Row():
with gr.Column(scale=1):
img_input = gr.Image(type="pil", label="Upload EO Image")
question_input = gr.Textbox(label="Ask a question about the image", placeholder="E.g. Where is the river?")
submit_btn = gr.Button("Analyze 🌟")
with gr.Column(scale=1):
output_text = gr.Textbox(label="AI Answer", interactive=False)
submit_btn.click(analyze_eo_image, inputs=[img_input, question_input], outputs=output_text)
with gr.Tab("Use Image URL"):
with gr.Row():
with gr.Column(scale=1):
url_input = gr.Textbox(label="Enter Image URL")
url_question = gr.Textbox(label="Ask a question about the image")
url_btn = gr.Button("Analyze 🌟")
with gr.Column(scale=1):
url_output = gr.Textbox(label="AI Answer", interactive=False)
url_btn.click(analyze_eo_url, inputs=[url_input, url_question], outputs=url_output)
gr.Markdown(
"💡 Tip: Use clear, simple questions for best results. Supports natural language queries about EO images."
)
demo.launch(share=True)
from huggingface_hub import login
login(token="hf_xxxx")
|