mujtaba_object_detection_code_export.py

# 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")