Upload app.py

app.py

@@ -0,0 +1,154 @@
+from IPython.display import display, JSON
+import matplotlib.pyplot as plt
+from speciesnet import DEFAULT_MODEL, SUPPORTED_MODELS, SpeciesNet
+import numpy as np
+import time
+import gradio as gr
+import json
+import cv2
+
+
+# --- Load SpeciesNet model ---
+print("Default SpeciesNet model:", DEFAULT_MODEL)
+print("Supported SpeciesNet models:", SUPPORTED_MODELS)
+model = SpeciesNet(DEFAULT_MODEL)
+
+
+# --- Visualization Function ---
+def draw_predictions(image_path, predictions_dict):
+    # Load image with OpenCV
+    img = cv2.imread(image_path)
+    if img is None:
+        raise ValueError(f"Could not load image: {image_path}")
+
+    img_h, img_w, _ = img.shape
+
+    for pred in predictions_dict.get("predictions", []):
+        detections = pred.get("detections", [])
+        classifications = pred.get("classifications", {})
+
+        # Get class names and scores
+        classes = classifications.get("classes", [])
+        scores = classifications.get("scores", [])
+
+        # Pick top classification if available
+        top_class_name = None
+        top_score = None
+        if len(classes) > 0:
+            top_class_name = classes[0].split(";")[-1]  # readable species name
+            top_score = scores[0]
+
+        # --- Draw detections ---
+        for i, det in enumerate(detections):
+            bbox = det["bbox"]  # [x, y, w, h] (relative coords)
+            conf = det["conf"]
+            label = det["label"]
+
+            # Convert relative bbox → pixel coordinates
+            x, y, w, h = bbox
+            x1 = int(x * img_w)
+            y1 = int(y * img_h)
+            x2 = int((x + w) * img_w)
+            y2 = int((y + h) * img_h)
+
+            # Draw bounding box
+            cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 3)
+
+            # Build text lines
+            detection_text = f"{label} ({conf:.2f})"
+            classification_text = (
+                f"{top_class_name} ({top_score:.2f})" if top_class_name else ""
+            )
+
+            # Combine both labels stacked vertically
+            text_lines = []
+            if classification_text:
+                text_lines.append(classification_text)
+            text_lines.append(detection_text)
+
+            # Compute text height for background box
+            total_text_height = 0
+            text_widths = []
+            for line in text_lines:
+                (text_w, text_h), _ = cv2.getTextSize(
+                    line, cv2.FONT_HERSHEY_SIMPLEX, 0.6, 2
+                )
+                total_text_height += text_h + 5
+                text_widths.append(text_w)
+
+            max_text_width = max(text_widths)
+
+            # Background for stacked labels
+            cv2.rectangle(
+                img,
+                (x1, max(y1 - total_text_height - 10, 0)),
+                (x1 + max_text_width + 10, y1),
+                (0, 255, 0),
+                -1,
+            )
+
+            # Write both lines
+            y_text = y1 - 5
+            for line in text_lines[::-1]:  # classification above detection
+                cv2.putText(
+                    img,
+                    line,
+                    (x1 + 5, y_text),
+                    cv2.FONT_HERSHEY_SIMPLEX,
+                    0.6,
+                    (0, 0, 0),
+                    2,
+                    cv2.LINE_AA,
+                )
+                (_, text_h), _ = cv2.getTextSize(line, cv2.FONT_HERSHEY_SIMPLEX, 0.6, 2)
+                y_text -= text_h + 5
+
+    # Convert BGR → RGB for display
+    img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    return img_rgb
+
+
+# --- Inference Function ---
+def inference(image):
+    filepath = "temp_image.jpg"
+    image.save(filepath)
+
+    start_time = time.time()
+    predictions_dict = model.predict(
+        instances_dict={
+            "instances": [
+                {
+                    "filepath": filepath,
+                    "country": "VNM",  # optional
+                }
+            ]
+        }
+    )
+    end_time = time.time()
+    print(f"Inference Time: {end_time - start_time:.2f} seconds")
+
+    # Draw predictions with detection + classification stacked
+    annotated_image = draw_predictions(filepath, predictions_dict)
+
+    # Return image + JSON
+    formatted_json = json.dumps(predictions_dict, indent=4)
+    return annotated_image, formatted_json
+
+
+# --- Gradio Interface ---
+iface = gr.Interface(
+    fn=inference,
+    inputs=gr.Image(type="pil"),
+    outputs=[
+        gr.Image(label="Detection + Classification Visualization"),
+        gr.JSON(label="Prediction Details"),
+    ],
+    title="🐾 SpeciesNet Wildlife Detector + Classifier",
+    description=(
+        "Upload a wildlife camera image. The model detects animals and shows both "
+        "the detection label (e.g., 'animal 0.97') and classification result "
+        "(e.g., 'white-tailed deer 0.99') on each bounding box."
+    ),
+)
+
+iface.launch()