Building a Brand Detection Model with YOLOv11 in Google Colab

📦 Building a Custom Object Detection Model (Brand Detection) with YOLOv11 in Google Colab 

Object detection has become significantly more accessible thanks to modern architectures like YOLO (You Only Look Once). This article walks through the entire process of setting up and training a YOLO model from scratch using custom image dataset — right inside Google Colab.

📁 1. Uploading and Preparing the Dataset

▶ Mount Google Drive (if your dataset is there)

from google.colab import drive

drive.mount('/content/gdrive')

This gives Colab access to your Google Drive so it can copy your dataset stored there.

▶ Copy the dataset ZIP file from Google Drive

!cp /content/gdrive/MyDrive/yolo/data.zip /content

Copies the ZIP archive containing your dataset into the Colab runtime.

▶ Alternative: Download from a URL

!wget -O /content/data.zip https://drive.google.com/file/d/1Hdrk5_dnJvJPoupPsJ0PaRyLPPDP7b4f/view?usp=sharing/data.zip

Downloads the dataset from a public URL directly into the Colab environment.

📦 2. Preparing the Data for Training

▶ Unzip the Dataset

!unzip -q /content/data.zip -d /content/custom_data

Extracts the dataset contents into a custom folder named (custom_data).

▶ Create Train and Validation Split Automatically

!wget -O /content/train_val_split.py 
https://drive.google.com/file/d/1Hdrk5_dnJvJPoupPsJ0PaRyLPPDP7b4f/view?usp=sharing/
train_val_split.py

Downloads a Python script that splits your dataset into 90% training and 10% validation 
images, in the correct folder structure expected by YOLO.


⚙️ 3. Install YOLO Training Requirements
!pip install ultralytics


Installs the official ultralytics package, which includes support for YOLOv5, 
YOLOv8, and YOLOv11.


🛠 4. Create the Configuration File (data.yaml)
▶ Why data.yaml?
YOLO models need a data.yaml file that tells the trainer:
Where to find training and validation images
How many classes there are
What the class names are
▶ Code to Auto-Generate the YAML File
import yaml
import os

def create_data_yaml(path_to_classes_txt, path_to_data_yaml):

  # Read class.txt to get class names
  if not os.path.exists(path_to_classes_txt):
    print(f'classes.txt file not found! Please create a classes.txt labelmap 
    and move it to {path_to_classes_txt}')
    return
  with open(path_to_classes_txt, 'r') as f:
    classes = []
    for line in f.readlines():
      if len(line.strip()) == 0: continue
      classes.append(line.strip())
  number_of_classes = len(classes)

  # Create data dictionary
  data = {
      'path': '/content/data',
      'train': 'train/images',
      'val': 'validation/images',
      'nc': number_of_classes,
      'names': classes
  }

  # Write data to YAML file
  with open(path_to_data_yaml, 'w') as f:
    yaml.dump(data, f, sort_keys=False)
  print(f'Created config file at {path_to_data_yaml}')

  return

# Define path to classes.txt and run function
path_to_classes_txt = '/content/custom_data/classes.txt'
path_to_data_yaml = '/content/data.yaml'

create_data_yaml(path_to_classes_txt, path_to_data_yaml)


This function:
Reads class names from classes.txt
Builds a dictionary with paths and labels
Saves it as data.yaml
▶ Run the Function
create_data_yaml('/content/custom_data/classes.txt', '/content/data.yaml')



🧠 5. Configure Training Settings
You can control:
Model size: (yolo11s.pt, yolo11m.pt, etc.)
Epochs: How many times to loop through training data
Resolution: Image size (e.g. 640x640)

!yolo detect train data=/content/data.yaml model=yolo11s.pt epochs=60 imgsz=640


💡 Note: Larger models like yolo11xl.pt offer better accuracy but train slower. 
For starters, yolo11s.pt is a good balance.


📊 6. Test the Trained Model
Once training is done, the best weights are automatically saved.
!yolo detect predict model=runs/detect/train/weights/best.pt 
source=data/validation/images save=True

This runs inference on your validation set using the trained model and saves predictions.

▶ Display Results in Notebook
import glob
from IPython.display import Image, display

for image_path in glob.glob('/content/runs/detect/predict/*.jpg')[:10]:
    display(Image(filename=image_path, height=400))


📷 7. Run Prediction on New Images
▶ Download a test image
!wget -O /content/data/test/images/test.jpg 
"https://images-cdn.ubuy.co.in/63526eb9c797ce1c1e1f26b4-snickers-m-amp-m-39-s.jpg"


▶ Predict and Show Result
!yolo detect predict model=runs/detect/train/weights/best.pt source=data/test/images save=True


from IPython.display import Image, display
display(Image(filename='/content/runs/detect/predict2/test.jpg', height=400))

This will show bounding boxes and labels on the new image using the trained model.