Flood-Image-Detection

Flood-Image-Detection is a vision-language encoder model fine-tuned from google/siglip2-base-patch16-512 for binary image classification. It is trained to detect whether an image contains a flooded scene or non-flooded environment. The model uses the SiglipForImageClassification architecture.

[!note] SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features : https://arxiv.org/pdf/2502.14786

Classification Report:
               precision    recall  f1-score   support

Flooded Scene     0.9172    0.9458    0.9313       609
  Non Flooded     0.9744    0.9603    0.9673      1309

     accuracy                         0.9557      1918
    macro avg     0.9458    0.9530    0.9493      1918
 weighted avg     0.9562    0.9557    0.9559      1918

Label Space: 2 Classes

Class 0: Flooded Scene  
Class 1: Non Flooded

Install Dependencies

pip install -q transformers torch pillow gradio hf_xet

Inference Code

import gradio as gr
from transformers import AutoImageProcessor, SiglipForImageClassification
from PIL import Image
import torch

# Load model and processor
model_name = "prithivMLmods/flood-image-detection"  # Update with actual model name on Hugging Face
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

# Updated label mapping
id2label = {
    "0": "Flooded Scene",
    "1": "Non Flooded"
}

def classify_image(image):
    image = Image.fromarray(image).convert("RGB")
    inputs = processor(images=image, return_tensors="pt")

    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        probs = torch.nn.functional.softmax(logits, dim=1).squeeze().tolist()

    prediction = {
        id2label[str(i)]: round(probs[i], 3) for i in range(len(probs))
    }

    return prediction

# Gradio Interface
iface = gr.Interface(
    fn=classify_image,
    inputs=gr.Image(type="numpy"),
    outputs=gr.Label(num_top_classes=2, label="Flood Detection"),
    title="Flood-Image-Detection",
    description="Upload an image to detect whether the scene is flooded or not."
)

if __name__ == "__main__":
    iface.launch()

Intended Use

Flood-Image-Detection is designed for:

• Disaster Monitoring – Rapid detection of flood-affected areas from imagery.
• Environmental Analysis – Track flooding patterns across regions using image datasets.
• Crisis Response – Assist emergency services in identifying critical zones.
• Surveillance and Safety – Monitor infrastructure or locations for flood exposure.
• Smart Alert Systems – Integrate with IoT or camera feeds for automated flood alerts.