Face-Confidence-SigLIP2(Experimental)

Face-Confidence-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 confident faces and unconfident faces using the SiglipForImageClassification architecture.

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

   confident     0.8468    0.8179    0.8321      4872
 unconfident     0.8691    0.8909    0.8799      6611

    accuracy                         0.8600     11483
   macro avg     0.8580    0.8544    0.8560     11483
weighted avg     0.8596    0.8600    0.8596     11483

Label Space: 2 Classes

The model classifies each image into one of the following categories:

Class 0: "confident"
Class 1: "unconfident"

---

Install Dependencies

pip install -q transformers torch pillow gradio

---

Image Scale (Optimal): 256 × 256

Inference Code

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

# Load model and processor
model_name = "prithivMLmods/Face-Confidence-SigLIP2"  # Replace with your model path if different
model = SiglipForImageClassification.from_pretrained(model_name)
processor = AutoImageProcessor.from_pretrained(model_name)

# Label mapping
id2label = {
    "0": "confident",
    "1": "unconfident"
}

def classify_face_confidence(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_face_confidence,
    inputs=gr.Image(type="numpy"),
    outputs=gr.Label(num_top_classes=2, label="Face Confidence Classification"),
    title="Face-Confidence-SigLIP2",
    description="Upload an image to detect if a face looks confident or unconfident."
)

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

---

Demo Inference(Image)

Intended Use

Face-Confidence-SigLIP2 can be used for:

• Behavioral Analysis – Detect confidence levels in facial expressions.
• Education & Training – Assess learner engagement or self-confidence.
• HR & Recruitment – Analyze non-verbal cues during interviews.
• Dataset Curation – Separate confident vs unconfident facial images for training.