Create app.py

app.py

@@ -0,0 +1,114 @@
+import gradio as gr
+from transformers import AutoModelForCausalLM, AutoProcessor
+import torch
+import numpy as np
+import cv2
+import logging
+
+# Logging einrichten
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+# Lade das Modell und den Processor
+model = AutoModelForCausalLM.from_pretrained("microsoft/florence-2-large", trust_remote_code=True)
+processor = AutoProcessor.from_pretrained("microsoft/florence-2-large", trust_remote_code=True)
+
+def analyze_image(image, prompt):
+    logger.info("Starting image analysis with prompt: %s", prompt)
+    
+    # Konvertiere PIL-Bild zu numpy-Format
+    image_np = np.array(image)
+    image_cv = cv2.cvtColor(image_np, cv2.COLOR_RGB2BGR)
+    logger.info("Image shape: %s", image_np.shape)
+
+    # Allgemeine Bildbeschreibung
+    if "what do you see" in prompt.lower() or "was siehst du" in prompt.lower():
+        inputs = processor(text=prompt, images=image_np, return_tensors="pt")
+        with torch.no_grad():
+            outputs = model.generate(
+                input_ids=inputs["input_ids"],
+                pixel_values=inputs["pixel_values"],
+                max_length=1024,
+                num_beams=3
+            )
+        description = processor.batch_decode(outputs, skip_special_tokens=True)[0]
+        return {"prompt": prompt, "description": description}
+
+    # Spezifische Kerzen-Analyse
+    elif "last 8 candles" in prompt.lower() or "letzte 8 kerzen" in prompt.lower():
+        # Objekterkennung mit Florence-2
+        task_prompt = "<OD>"  # Objekterkennung
+        inputs = processor(text=task_prompt, images=image_np, return_tensors="pt")
+        with torch.no_grad():
+            outputs = model.generate(
+                input_ids=inputs["input_ids"],
+                pixel_values=inputs["pixel_values"],
+                max_length=1024,
+                num_beams=3
+            )
+        predictions = processor.post_process_generation(outputs, task=task_prompt, image_size=(image_np.shape[1], image_np.shape[0]))
+        logger.info("Detected objects: %s", predictions)
+
+        # Extrahiere Kerzen
+        detections = []
+        if "<OD>" in predictions:
+            for i, (bbox, label) in enumerate(zip(predictions["<OD>"]["bboxes"], predictions["<OD>"]["labels"])):
+                if "candle" not in label.lower():  # Filtere nur Kerzen
+                    continue
+                xmin, ymin, xmax, ymax = map(int, bbox)
+                
+                # Extrahiere Farbe
+                candle_roi = image_cv[ymin:ymax, xmin:xmax]
+                if candle_roi.size == 0:
+                    logger.warning("Empty ROI for box: (%d, %d, %d, %d)", xmin, ymin, xmax, ymax)
+                    continue
+                mean_color = np.mean(candle_roi, axis=(0, 1)).astype(int)
+                color_rgb = f"RGB({mean_color[2]},{mean_color[1]},{mean_color[0]})"
+
+                # OCR für Preise (Achsen oder Kerzenregion)
+                price_roi = image_cv[max(0, ymin-50):min(image_np.shape[0], ymax+50), 
+                                    max(0, xmin-50):min(image_np.shape[1], xmax+50)]
+                price_task = "<OCR>"
+                ocr_inputs = processor(text=price_task, images=price_roi, return_tensors="pt")
+                with torch.no_grad():
+                    ocr_outputs = model.generate(
+                        input_ids=ocr_inputs["input_ids"],
+                        pixel_values=ocr_inputs["pixel_values"],
+                        max_length=1024
+                    )
+                prices = processor.batch_decode(ocr_outputs, skip_special_tokens=True)[0]
+
+                detections.append({
+                    "pattern": label,
+                    "color": color_rgb,
+                    "prices": prices if prices else "No price detected",
+                    "x_center": (xmin + xmax) / 2
+                })
+
+        # Sortiere nach x-Position (rechts nach links = neueste Kerzen)
+        detections = sorted(detections, key=lambda x: x["x_center"], reverse=True)[:8]
+        logger.info("Sorted detections: %d", len(detections))
+
+        if not detections:
+            logger.warning("No candlesticks detected. Ensure clear image with visible candles.")
+            return {"prompt": prompt, "description": "No candlesticks detected. Try a clearer screenshot."}
+
+        return {"prompt": prompt, "detections": detections}
+
+    # Fallback für unbekannte Prompts
+    else:
+        return {"prompt": prompt, "description": "Unsupported prompt. Use 'Was siehst du auf dem Bild?' or 'List last 8 candles with their colors'."}
+
+# Erstelle Gradio-Schnittstelle
+iface = gr.Interface(
+    fn=analyze_image,
+    inputs=[
+        gr.Image(type="pil", label="Upload an Image"),
+        gr.Textbox(label="Prompt", placeholder="Enter your prompt, e.g., 'Was siehst du auf dem Bild?' or 'List last 8 candles with their colors'")
+    ],
+    outputs="json",
+    title="Image Analysis with Florence-2-large",
+    description="Upload an image and provide a prompt to get a description or analyze candlesticks."
+)
+
+iface.launch()