feat: Add Hugging Face Spaces demo link, remove Hugging Face login requirement for dataset download, and detail the computational environment.

README.md

@@ -14,14 +14,17 @@ license: mit
 [![Python 3.10+](https://img.shields.io/badge/python-3.10+-blue.svg)](https://www.python.org/downloads/)
 [![Anomalib](https://img.shields.io/badge/anomalib-2.2.0-green.svg)](https://github.com/openvinotoolkit/anomalib)
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
+[![Hugging Face Spaces](https://img.shields.io/badge/🤗%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/micguida1/mvtec-anomaly-benchmark)
 
 A comprehensive benchmark for anomaly detection models on the [MVTec AD dataset](https://www.mvtec.com/company/research/datasets/mvtec-ad) using [Anomalib](https://github.com/openvinotoolkit/anomalib).
 
+**🚀 [Try the Live Demo on Hugging Face Spaces](https://huggingface.co/spaces/micguida1/mvtec-anomaly-benchmark)**
+
 ## 🎯 Features
 
 - **Multiple Models**: PatchCore, EfficientAD, FastFlow, STFPM, PaDiM
 - **Full Benchmark**: Train and evaluate on all 15 MVTec categories
-- **Interactive Demo**: Gradio UI for real-time anomaly detection
+- **Interactive Demo**: [Gradio UI for real-time anomaly detection](https://huggingface.co/spaces/micguida1/mvtec-anomaly-benchmark)
 - **Easy Configuration**: YAML-based model configs
 
 ## 📦 Installation
@@ -50,8 +53,8 @@ python scripts/download_mvtec.py
 
 The script features an **interactive menu** where you can choose between:
 
-1.  **Hugging Face** (Recommended - Fast): Downloads from `micguida1/mvtech_anomaly_detection`. Requires a Hugging Face account. The script will guide you through the login process if needed.
-2.  **HTTP Mirror** (Fallback): Downloads from the original public mirror (~5GB, slower). No login required.
+1.  **Hugging Face** (Recommended - Fast): Downloads from `micguida1/mvtech_anomaly_detection`. No login required.
+2.  **HTTP Mirror** (Fallback): Downloads from the original public mirror (~5GB, slower).
 
 The dataset will be automatically extracted to `data/MVTecAD/`.
 
@@ -169,3 +172,15 @@ This project is licensed under the MIT License - see the [LICENSE](LICENSE) file
 - [Anomalib](https://github.com/openvinotoolkit/anomalib) - Anomaly detection library
 - [MVTec AD](https://www.mvtec.com/company/research/datasets/mvtec-ad) - Dataset
 
+## 💻 Computational Environment
+
+All experiments were conducted on a cloud machine rented via [Lightning.ai](https://lightning.ai/) with the following specifications:
+
+| Component | Specification |
+|-----------|---------------|
+| **CPU** | Intel® Xeon® Platinum 8468 (16 vCPUs, 8 physical cores @ 2.1 GHz) |
+| **RAM** | 196 GB |
+| **GPU** | NVIDIA H200 (141 GB HBM3 VRAM) |
+
+This high-performance setup enabled fast training and evaluation of all models across the entire MVTec AD dataset.
+

scripts/download_mvtec.py

@@ -29,8 +29,8 @@ def download_mvtec():
 
     print("\n--- BANCA DATI MVTEC AD - DOWNLOADER ---")
     print("Scegli il metodo di download:")
-    print("1. Hugging Face (Veloce - Richiede Login)")
-    print("2. HTTP Mirror (Lento ~5GB - No Login)")
+    print("1. Hugging Face (Recommended - Fast)")
+    print("2. HTTP Mirror (Slower ~5GB)")
     print("q. Esci")
     
     while True:
@@ -53,51 +53,32 @@ def download_mvtec():
         extract_dataset(target_dir, archive_name)
 
 def download_huggingface(target_dir, archive_name):
-    print("\n--- Metodo: Hugging Face ---")
+    print("\n--- Method: Hugging Face ---")
     try:
-        from huggingface_hub import hf_hub_download, login
+        from huggingface_hub import hf_hub_download
     except ImportError:
-        print("Libreria 'huggingface_hub' non trovata. Installing...")
+        print("Library 'huggingface_hub' not found. Installing...")
         subprocess.check_call([sys.executable, "-m", "pip", "install", "huggingface_hub"])
-        from huggingface_hub import hf_hub_download, login
+        from huggingface_hub import hf_hub_download
 
-    print("Tentativo di accesso...")
     try:
-        # Verifica accesso provando a scaricare info (o file piccolo/header)
-        # Usiamo force_download=False per check veloce, ma se fallisce gestiamo
-        hf_hub_download(
-            repo_id="micguida1/mvtech_anomaly_detection",
-            filename="mvtec_anomaly_detection.tar.xz",
-            repo_type="dataset",
-            local_dir=target_dir,
-            local_dir_use_symlinks=False
-        )
-        print("Login valido o cache presente.")
-    except Exception as e:
-        print(f"\nAccesso non riuscito o file non trovato ({e}).")
-        print("È necessario il login a Hugging Face (se il repo è privato).")
-        print("Inserisci il tuo token (lo trovi su https://huggingface.co/settings/tokens)")
-        login()
-    
-    try:
-        print("Avvio download da Hugging Face (repo: micguida1/mvtech_anomaly_detection)...")
+        print("Starting download from Hugging Face (repo: micguida1/mvtech_anomaly_detection)...")
         filepath = hf_hub_download(
             repo_id="micguida1/mvtech_anomaly_detection",
             filename="mvtec_anomaly_detection.tar.xz",
             repo_type="dataset",
             local_dir=target_dir,
-            local_dir_use_symlinks=False,
-            force_download=True # Forza riscaricamento per evitare file corrotti in cache
+            local_dir_use_symlinks=False
         )
-        # Fix path se necessario
+        # Fix path if needed
         if os.path.exists(filepath) and filepath != archive_name:
             if os.path.exists(archive_name):
                 os.remove(archive_name)
             os.rename(filepath, archive_name)
-        print("Download completato!")
+        print("Download completed!")
         return True
     except Exception as e:
-        print(f"Errore download HF: {e}")
+        print(f"HF download error: {e}")
         return False
 
 def download_http(archive_name):