File size: 3,928 Bytes
e73d9f4
 
 
b96d0e2
 
 
e73d9f4
 
 
 
 
 
 
 
 
 
b96d0e2
 
e73d9f4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b96d0e2
 
e73d9f4
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131




# these are the libraries you have to isntall for this projet 
#!pip install ultralytics datasets wandb gradio opencv-python Pillow captum torchvision --upgrade

import os
import cv2
import torch
import numpy as np
from ultralytics import YOLO
import wandb
import matplotlib.pyplot as plt
from datetime import datetime
from google.colab import userdata
# below is the api key's name I had set in my google colab's secret key name to ensure privacy for my api key . If you want 
# you can create an accoint at WANDB and then download name the api key WANDB and then use it for your personal use
wandb.login(key=userdata.get('WANDB'))

def setup_wandb():
    wandb.init(project="Object-detection",
              name=f"run_{datetime.now().strftime('%Y%m%d_%H%M%S')}",
              config={
                  "model": "yolov8n",
                  "dataset": "coco128",
                  "img_size": 640,
                  "batch_size": 8
              })

def load_model():
    model = YOLO("yolov8n.pt")
    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    model.to(device)
    return model

def train_model(model):
    results = model.train(
        data="coco128.yaml",
        epochs=20,
        imgsz=640,
        batch=8,
        device='0' if torch.cuda.is_available() else 'cpu',
        patience=3,
        save=True
    )
    return model

def validate_model(model):
    metrics = model.val()
    wandb.log({
        "val/mAP50": metrics.box.map50,
        "val/mAP50-95": metrics.box.map,
        "val/precision": metrics.box.mp,
        "val/recall": metrics.box.mr
    })
    return metrics

def visualize_results(results, img_path):
    img = cv2.imread(img_path)
    if img is None:
        raise ValueError(f"Failed to load image: {img_path}")
    pred_img = results[0].plot()
    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(20, 10))
    ax1.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
    ax1.axis('off')
    ax2.imshow(cv2.cvtColor(pred_img, cv2.COLOR_BGR2RGB))
    ax2.axis('off')
    plt.savefig("detection_results.jpg")
    plt.close()
    return "detection_results.jpg"

def test_image(model, img_path="test_image.jpg"):
    if not os.path.exists(img_path):
        raise FileNotFoundError(f"Image not found: {img_path}")
    results = model(img_path)
    output_path = visualize_results(results, img_path)
    wandb.log({
        "test_results": wandb.Image(output_path),
        "detections": results[0].boxes.cls.tolist(),
        "confidences": results[0].boxes.conf.tolist()
    })
    return results

def webcam_demo(model):
    try:
        from google.colab.patches import cv2_imshow
        cap = cv2.VideoCapture(0)
        if not cap.isOpened():
            print("Webcam not available - skipping demo")
            return
        print("Press 'q' to quit webcam demo")
        while True:
            ret, frame = cap.read()
            if not ret:
                break
            results = model(frame)
            annotated = results[0].plot()
            cv2_imshow(annotated)
            if cv2.waitKey(1) & 0xFF == ord('q'):
                break
    except Exception as e:
        print(f"Webcam error: {e}")
    finally:
        cap.release()
        cv2.destroyAllWindows()

def export_model():
    trained_weights = "runs/detect/train/weights/best.pt"
    model = YOLO(trained_weights)
    model.export(format="torchscript")
    wandb.save("best.torchscript")

def main():
    setup_wandb()
    model = load_model()
    model = train_model(model)
    validate_model(model)
    test_image(model)
    export_model()
    wandb.finish()

if __name__ == "__main__":
    main()
# main objective of this colab filw was to generate wegihts after training and thne later downlaod and upload to hugging face for inference
#as you cannot directly run of hugging due to absence of gpu
from google.colab import files

files.download("runs/detect/train/weights/best.torchscript")