Upload 7 files

.gitattributes

@@ -1,35 +1 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,12 +1,28 @@
----
-title: Resnet34 Testrun
-emoji: ⚡
-colorFrom: red
-colorTo: indigo
-sdk: gradio
-sdk_version: 5.45.0
-app_file: app.py
-pinned: false
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+# ResNet34 Corrosion Classifier — Hugging Face Space
+
+Semplice Space Gradio che carica un modello ResNet34 e predice 9 classi di corrosione.
+
+## File
+- `app.py`: interfaccia Gradio.
+- `model.py`: definizione modello e caricamento pesi.
+- `classes.json`: etichette delle classi.
+- `requirements.txt`: dipendenze.
+- `.gitattributes`: abilita LFS per i file `.pth`.
+- `resnet34_best.pth`: **DA CARICARE DA TE** (non incluso).
+
+## Istruzioni
+1. Crea una nuova Space su Hugging Face (Gradio + Python).
+2. Carica questi file nella Space.
+3. Aggiungi il tuo file di pesi `resnet34_best.pth` (usa Git LFS se > 50 MB).
+4. (Opzionale) Se il file si chiama diversamente, imposta una variabile d'ambiente `CKPT_PATH`
+   nelle Settings della Space, oppure modifica `CKPT_PATH` in `app.py`.
+5. Avvia la Space.
+
+## Uso
+- Carica o scatta una foto, poi clicca **Analizza immagine**.
+- La card a destra mostra le probabilità (Top-K) e la predizione.
+
+## Note
+- Il modello gira su CPU per default. Se vuoi più velocità, passa a una Space con GPU.
+- Le trasformazioni input usano Resize 256 → CenterCrop 224 e normalizzazione ImageNet.

app.py

@@ -0,0 +1,82 @@
+
+import os
+import json
+import torch
+import torch.nn.functional as F
+from PIL import Image
+from torchvision import transforms
+import gradio as gr
+
+from model import build_model, load_weights
+
+TITLE = "ResNet34 Corrosion Classifier"
+DESCRIPTION = \"\"\
+Carica o scatta una foto. Il modello (ResNet34) restituisce la classe prevista e le probabilità.
+Assicurati di **caricare il file dei pesi** nella repo come `resnet34_best.pth` (o aggiorna il percorso qui sotto).
+\"\"\
+
+# ====== Config ======
+CKPT_PATH = os.environ.get("CKPT_PATH", "resnet34_best.pth")
+CLASSES_PATH = os.environ.get("CLASSES_PATH", "classes.json")
+DEVICE = "cpu"  # su Spaces CPU per default
+
+with open(CLASSES_PATH, "r", encoding="utf-8") as f:
+    IDX2LABEL = json.load(f)
+
+preprocess = transforms.Compose([
+    transforms.Resize(256, interpolation=transforms.InterpolationMode.BICUBIC),
+    transforms.CenterCrop(224),
+    transforms.ToTensor(),
+    transforms.Normalize(mean=[0.485, 0.456, 0.406],
+                         std=[0.229, 0.224, 0.225]),
+])
+
+# Lazy load del modello
+_model = None
+def get_model():
+    global _model
+    if _model is None:
+        model = build_model(num_classes=len(IDX2LABEL))
+        if not os.path.isfile(CKPT_PATH):
+            raise FileNotFoundError(
+                f\"Checkpoint non trovato: {CKPT_PATH}. Carica i pesi nella Space o imposta CKPT_PATH.\"
+            )
+        model = load_weights(model, CKPT_PATH, map_location=DEVICE)
+        _model = model
+    return _model
+
+def predict(image: Image.Image, topk: int = 5):
+    if image is None:
+        return {"Errore": 1.0}, "Nessuna immagine."
+    model = get_model()
+    model.eval()
+    with torch.no_grad():
+        img = image.convert("RGB")
+        tensor = preprocess(img).unsqueeze(0)
+        logits = model(tensor)
+        probs = F.softmax(logits, dim=1).squeeze(0)
+        topk = min(topk, probs.shape[0])
+        values, indices = torch.topk(probs, k=topk)
+        label_scores = {IDX2LABEL[i.item()]: float(v.item()) for v, i in zip(values, indices)}
+        pred_label = IDX2LABEL[int(torch.argmax(probs).item())]
+        msg = f"Predizione: **{pred_label}**"
+        return label_scores, msg
+
+with gr.Blocks(fill_height=True) as demo:
+    gr.Markdown(f"# {TITLE}")
+    gr.Markdown(DESCRIPTION)
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            img_in = gr.Image(type="pil", sources=["upload", "webcam"], label="Immagine")
+            topk = gr.Slider(1, 9, value=5, step=1, label="Top-K")
+            btn = gr.Button("Analizza immagine")
+        with gr.Column(scale=1):
+            lbl = gr.Label(label="Probabilità", num_top_classes=9)
+            txt = gr.Markdown()
+
+    btn.click(predict, inputs=[img_in, topk], outputs=[lbl, txt])
+    img_in.change(predict, inputs=[img_in, topk], outputs=[lbl, txt])
+
+if __name__ == "__main__":
+    demo.launch()

classes.json

@@ -0,0 +1,11 @@
+[
+  "crevice_corrosion",
+  "erosion_corrosion",
+  "galvanic_corrosion",
+  "mic_corrosion",
+  "no_corrosion",
+  "pitting_corrosion",
+  "stress_corrosion",
+  "under_insulation_corrosion",
+  "uniform_corrosion"
+]

model.py

@@ -0,0 +1,28 @@
+
+import torch
+import torch.nn as nn
+from torchvision import models
+
+def build_model(num_classes: int) -> nn.Module:
+    model = models.resnet34(weights=None)
+    in_features = model.fc.in_features
+    model.fc = nn.Linear(in_features, num_classes)
+    return model
+
+def load_weights(model: nn.Module, ckpt_path: str, map_location="cpu") -> nn.Module:
+    state = torch.load(ckpt_path, map_location=map_location)
+    # Support both full state dicts and {'model': state_dict} formats
+    if isinstance(state, dict) and "state_dict" in state:
+        state = state["state_dict"]
+    if isinstance(state, dict) and "model" in state and isinstance(state["model"], dict):
+        state = state["model"]
+    # Strip possible 'module.' prefixes
+    new_state = {}
+    for k, v in state.items():
+        if k.startswith("module."):
+            new_state[k[len("module."):]] = v
+        else:
+            new_state[k] = v
+    model.load_state_dict(new_state, strict=False)
+    model.eval()
+    return model

requirements.txt

@@ -0,0 +1,5 @@
+torch>=2.2.0
+torchvision>=0.17.0
+pillow>=10.3.0
+numpy>=1.26.4
+gradio==4.44.1

resnet34_best.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f03fbf24fb21f3eb66b8d348bbf908408c0c3b4176384ddc341ad23da3d553a
+size 255709619