---
title: Gambling Site Detector
emoji: 🐢
colorFrom: gray
colorTo: purple
sdk: gradio
sdk_version: 5.27.0
app_file: app.py
pinned: false
license: mit
short_description: Detects whether a website is related to gambling or not
---

# 🕵️ Indonesian Gambling Website Detection - Huggingface Space

This Space detects whether a website is related to gambling based on its screenshot and OCR text.

## Features
- Single URL Prediction
- Batch URLs Prediction
- Screenshot capture using external API (apiflash)
- OCR extraction using a hybrid approach (OCR.Space API + EasyOCR fallback)
- Multimodal model (image + text fusion)

## Instructions
1. Enter a website URL or upload a `.txt` file containing multiple URLs (one URL per line).
2. The system will:
   - Take a screenshot of the website.
   - Extract text using Tesseract OCR.
   - Predict if it is a Gambling or Non-Gambling site.

## Model
- Fusion model (`best_fusion_model.pth`) trained with EfficientNet + IndoBERT.
<!-- - Hosted at: [HuggingFace Model Hub](https://huggingface.co/azzandr/gambling-fusion-model) -->

## Deployment
This Space requires:
- Gradio
- Torch
- Transformers
- EasyOCR
- Pillow
- Pandas
- Requests

## Important Notes
⚡ **Inference may take longer than expected** because this Space runs on **CPU-only hardware**.  
Performance is significantly slower compared to GPU-enabled environments like Google Colab.  
Each prediction can take several minutes due to the complexity of multimodal fusion models (image + text processing).

For faster performance, consider running the model locally with a GPU, or upgrading to a GPU-enabled Huggingface Space.