Food-or-Not-SigLIP2
Food-or-Not-SigLIP2 is a vision-language encoder model fine-tuned from google/siglip2-base-patch16-224 for binary image classification. It is trained to distinguish between images of food and non-food objects using the SiglipForImageClassification architecture.
Classification Report:
precision recall f1-score support
food 0.8902 0.8610 0.8753 4000
not-food 0.8654 0.8938 0.8794 4000
accuracy 0.8774 8000
macro avg 0.8778 0.8774 0.8773 8000
weighted avg 0.8778 0.8774 0.8773 8000
Label Space: 2 Classes
The model classifies each image into one of the following categories:
Class 0: "food"
Class 1: "not-food"
Install Dependencies
pip install -q transformers torch pillow gradio
Inference Code
import gradio as gr
from transformers import AutoImageProcessor, SiglipForImageClassification
from PIL import Image
import torch
# Load model and processor
model_name = "prithivMLmods/Food-or-Not-SigLIP2" # Replace with your model path if different
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)
# Label mapping
id2label = {
"0": "food",
"1": "not-food"
}
def classify_food(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_food,
inputs=gr.Image(type="numpy"),
outputs=gr.Label(num_top_classes=2, label="Food Classification"),
title="Food-or-Not-SigLIP2",
description="Upload an image to detect if it contains food or not."
)
if __name__ == "__main__":
iface.launch()
Intended Use
Food-or-Not-SigLIP2 can be used for:
- Dietary Apps – Automatically classify images for food detection.
- Retail & E-commerce – Filter food vs non-food products visually.
- Content Moderation – Flag content containing food items.
- Dataset Curation – Separate food-related images for training or filtering.