Issue 6: Increase the Count of Less Represented Classes #102
Description
Problem: Class Imbalance in the Dataset
Some classes in the dataset have very few images compared to others. This imbalance can cause the model to focus more on classes with many images and perform poorly on the classes with fewer samples.
Goal
Increase the number of images in the less represented classes so that all classes are better balanced during training.
What Can Be Done
Create additional variations of existing images in the smaller classes by:
- Flipping images
- Rotating images
- Adjusting brightness and contrast
- Changing colors slightly
- Applying other small visual changes
These new images should be added only to the classes that have fewer samples.
📈 Outcome
After adding these variations, all classes will be better represented during training, which can help the model learn more fairly and improve overall performance. This process is called augmentation.
Its preferrable to apply augmentation only in train folder.
Contribution details:
Implementation Notes
->This task must be done inside the participants folder in your enrolment number's folder.
->You may Implement it in a separate notebook, or Add it to a previously used notebook.
->If working on Kaggle:
Make the required changes directly in your existing Kaggle notebook.
Download the updated notebook after changes.
Upload the updated version to the repository.
->Follow the PR template as given in the previous issues.