Upload 4 files

app.py

@@ -0,0 +1,46 @@
+import torch
+from torchvision import models, transforms
+from PIL import Image
+import gradio as gr
+
+# Modell laden
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+model = models.resnet18()
+model.fc = torch.nn.Linear(model.fc.in_features, 9)
+model.load_state_dict(torch.load("emotion_model.pt", map_location=device))
+model = model.to(device)
+model.eval()
+
+# Labels
+labels = ["happy", "sad", "angry", "surprised", "fear", "disgust", "neutral", "contempt", "unknown"]
+
+# Transformation
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor()
+])
+
+def predict_emotion(image):
+    image = image.convert("RGB")
+    image = transform(image).unsqueeze(0).to(device)
+
+    with torch.no_grad():
+        outputs = model(image)
+        probs = torch.softmax(outputs, dim=1)
+        confidence, predicted = torch.max(probs, 1)
+
+    if confidence.item() < 0.7:
+        return "Unbekannte Emotion", f"{confidence.item()*100:.2f}%"
+    else:
+        return labels[predicted.item()], f"{confidence.item()*100:.2f}%"
+
+# Gradio Interface
+interface = gr.Interface(
+    fn=predict_emotion,
+    inputs=gr.Image(type="pil"),
+    outputs=["text", "text"],
+    title="Emotion Recognition App",
+    description="Lade ein Bild hoch und erkenne die Emotion."
+)
+
+interface.launch()

emotion_model .pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71d4a41b5a1e1b3d14dd36e8db9e3773aa476891c9dfcc4c753912a25e7e6039
+size 44802354

requirements.txt

@@ -0,0 +1,4 @@
+gradio
+torch
+torchvision
+pillow

test_emotions_ood.py

@@ -0,0 +1,51 @@
+# test_emotions_ood.py
+
+import torch
+from torchvision import models, transforms
+from PIL import Image
+import os
+
+# Gerät wählen
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+# Modell definieren (gleiche Architektur wie beim Training)
+model = models.resnet18()
+model.fc = torch.nn.Linear(model.fc.in_features, 9)  # 9 Emotionen
+
+# Modellpfad und Laden
+model_path = "emotion_model.pt"
+model.load_state_dict(torch.load(model_path, map_location=device))
+model = model.to(device)
+model.eval()
+
+# Label Mapping (Anpassen an deine Klassen)
+labels = ["happy", "sad", "angry", "surprised", "fear", "disgust", "neutral", "contempt", "unknown"]
+
+# Transformation für Bilder
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor()
+])
+
+def predict_image(image_path, threshold=0.7):
+    image = Image.open(image_path).convert("RGB")
+    image = transform(image).unsqueeze(0).to(device)
+
+    with torch.no_grad():
+        outputs = model(image)
+        probs = torch.softmax(outputs, dim=1)
+        confidence, predicted = torch.max(probs, 1)
+
+    if confidence.item() < threshold:
+        return "Unbekannte Emotion erkannt", confidence.item()
+    else:
+        return labels[predicted.item()], confidence.item()
+
+# Beispielhafte Nutzung
+test_image_path = "testbild.jpg"  # <-- Dein Testbild hier
+
+if os.path.exists(test_image_path):
+    emotion, conf = predict_image(test_image_path)
+    print(f"Erkannte Emotion: {emotion} (Confidence: {conf:.2f})")
+else:
+    print(f"Bild {test_image_path} nicht gefunden.")