initial files added

app.py

@@ -0,0 +1,52 @@
+import gradio as gr
+from detection import ObjectDetection
+
+def get_predictions(img, threshold, box_color, text_color):
+    v3_results = yolov3_detector.score_frame(img)
+    v5_results = yolov5_detector.score_frame(img)
+    v8_results = yolov8_detector.v8_score_frame(img)
+
+    v3_frame = yolov3_detector.plot_bboxes(v3_results, img, float(threshold), box_color, text_color)
+    v5_frame = yolov5_detector.plot_bboxes(v5_results, img, float(threshold), box_color, text_color)
+    v8_frame = yolov8_detector.plot_bboxes(v8_results, img, float(threshold), box_color, text_color)
+
+
+    return v3_frame, v5_frame, v8_frame
+
+
+with gr.Blocks(title="Leaf Disease Detection", theme=gr.themes.Monochrome()) as interface:
+    gr.Markdown("# Leaf Disease Detection")
+    with gr.Row():
+        with gr.Column():
+            image = gr.Image(shape=(416,416), label="Input Image")
+        with gr.Column():
+            with gr.Row():
+                with gr.Column():
+                    box_color = gr.ColorPicker(label="Box Color", value="#0000ff")
+                with gr.Column():
+                    text_color = gr.ColorPicker(label="Prediction Color", value="#ff0000")
+
+            confidence = gr.Slider(maximum=1, step=0.01, value=0.4, label="Confidence Threshold", interactive=True)
+            btn = gr.Button("Detect")
+    
+    with gr.Row():        
+        with gr.Box():
+            v3_prediction = gr.Image(shape=(416,416), label="YOLOv3")
+        with gr.Box():   
+            v5_prediction = gr.Image(shape=(416,416), label="YOLOv5")
+        with gr.Box():   
+            v8_prediction = gr.Image(shape=(416,416), label="YOLOv8")
+
+        btn.click(
+            get_predictions,
+            [image, confidence, box_color, text_color],
+            [v3_prediction, v5_prediction, v8_prediction]
+        )
+
+
+yolov3_detector = ObjectDetection('yolov3')
+yolov5_detector = ObjectDetection('yolov5')
+yolov8_detector = ObjectDetection('yolov8')
+
+interface.launch()
+

detection.py

@@ -0,0 +1,82 @@
+import torch
+import cv2
+from PIL import ImageColor
+from ultralytics import YOLO
+
+
+class ObjectDetection:
+    def __init__(self, model_name='yolov3'):
+        self.model_name = model_name
+        self.model = self.load_model()
+        self.classes = self.model.names
+        self.device = 'cpu' 
+
+    def load_model(self):
+        if self.model_name != 'yolov8':
+            model = torch.hub.load(f"ultralytics/{self.model_name}", 'custom', path=f"weights/{self.model_name}_best.pt", force_reload=True)
+        else:
+            model = YOLO(f"weights/{self.model_name}_best.pt")
+
+        return model
+    
+    def score_frame(self, frame):
+        results = self.model(frame)
+
+        labels, conf, coord = results.xyxyn[0][:, -1], results.xyxyn[0][:, -2], results.xyxyn[0][:, :-1]
+        return labels, conf, coord
+
+    
+    def v8_score_frame(self, frame):
+        results = self.model(frame)
+
+        labels = []
+        confidences = []
+        coords = []
+
+        for result in results:
+            boxes = result.boxes.cpu().numpy()
+            
+            label = boxes.cls
+            conf = boxes.conf
+            coord = boxes.xyxy
+
+            labels.extend(label)
+            confidences.extend(conf)
+            coords.extend(coord)
+
+        return labels, confidences, coords
+
+    def get_coords(self, frame, row):
+
+        if self.model_name != 'yolov8':
+            x_shape, y_shape = frame.shape[1], frame.shape[0]
+            return int(row[0] * x_shape), int(row[1] * y_shape), int(row[2] * x_shape), int(row[3] * y_shape)
+        else:
+            return int(row[0]), int(row[1]), int(row[2]), int(row[3])
+    
+    def class_to_label(self, x):
+        return self.classes[int(x)]
+    
+    def get_color(self, code):
+        rgb = ImageColor.getcolor(code, "RGB")
+        return rgb 
+    
+    def plot_bboxes(self, results, frame, threshold, box_color, text_color):
+        labels, conf, coord = results
+        n = len(labels)
+
+        frame = frame.copy()
+        box_color = self.get_color(box_color)
+        text_color = self.get_color(text_color)
+
+        for i in range(n):
+            row = coord[i]
+            if conf[i] >= threshold:
+                x1, y1, x2, y2 = self.get_coords(frame, row)
+                class_name = self.class_to_label(labels[i])
+
+                cv2.rectangle(frame, (x1, y1), (x2, y2), box_color, 2)
+                cv2.putText(frame, f"{class_name} - {conf[i]*100:.2f}%", (x1, y1), cv2.FONT_HERSHEY_COMPLEX, 0.5, text_color)
+        
+        return frame
+

requirements.txt

@@ -0,0 +1,5 @@
+torch
+cv2
+PIL
+ultralytics
+gradio

weights/yolov3_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de10bcf127d161a673b004e0972c36af506299e14d261b9a0fcdb7ae9fbe77b0
+size 93039765

weights/yolov5_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77d41071f2ece081a264deac85cd7c6526c814627ef2e6c34137242947c3a33e
+size 93039637

weights/yolov8_best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daed75e7e8751a6a0a5e9aabc77556fa7367de4d4e2511448648232aeb43842e
+size 87651208