Delete inference.py

inference.py

@@ -1,68 +0,0 @@
-import torch
-from transformers import Pipeline
-from segment_anything import sam_model_registry, SamPredictor
-import numpy as np
-
-class CVATWrapper:
-    def __init__(self, model_path):
-        # Modell laden
-        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
-        
-        # Für SAM:
-        self.model = sam_model_registry["vit_h"](checkpoint=model_path)
-        self.model.to(self.device)
-        self.predictor = SamPredictor(self.model)
-        
-        # Für eigene Modelle:
-        # self.model = torch.load(model_path)
-        # self.model.eval()
-
-    def preprocess(self, image):
-        # Bildvorverarbeitung
-        if isinstance(image, str):
-            # Falls Bildpfad übergeben wird
-            image = cv2.imread(image)
-            image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
-        return image
-
-    def predict(self, image):
-        # Vorverarbeitung
-        image = self.preprocess(image)
-        
-        # Für SAM:
-        self.predictor.set_image(image)
-        masks, scores, logits = self.predictor.predict()
-        
-        # Für eigene Modelle:
-        # output = self.model(processed_image)
-        # masks = process_output(output)  # Je nach Modellausgabe
-        
-        # Konvertierung ins CVAT-Format
-        results = []
-        for mask, score in zip(masks, scores):
-            if score > 0.5:  # Konfidenz-Schwellenwert
-                results.append({
-                    "confidence": float(score),
-                    "label": "object",
-                    "points": mask_to_polygons(mask),  # Hilfsfunktion unten
-                    "type": "polygon"
-                })
-        
-        return results
-
-def mask_to_polygons(mask):
-    """Konvertiert eine binäre Maske in CVAT-Polygon-Format"""
-    import cv2
-    contours, _ = cv2.findContours(
-        mask.astype(np.uint8), 
-        cv2.RETR_EXTERNAL, 
-        cv2.CHAIN_APPROX_SIMPLE
-    )
-    polygons = []
-    for contour in contours:
-        # Vereinfache die Konturen
-        epsilon = 0.005 * cv2.arcLength(contour, True)
-        approx = cv2.approxPolyDP(contour, epsilon, True)
-        if len(approx) >= 3:  # Mindestens 3 Punkte für ein Polygon
-            polygons.append(approx.reshape(-1, 2).tolist())
-    return polygons[0] if polygons else []