Skip to main content
Hemanth.
Back to Feed
Jun 2024 – Jun 2024AI & Data

Image Classifier using Tensorflow

A comprehensive deep learning model building experience producing high validation accuracy by classifying images into organized folders.

Rationale

What I Learned:

  • The critical importance of balanced and well-preprocessed datasets.
  • Advanced techniques in CNN architecture design and hyperparameter tuning.
  • Practical implementation of machine learning models for real-world applications.

I loved learning new things natively manipulating models using Google Colab, Keras, and TensorFlow. Tracking epochs and tuning hyperparameters in loops was a challenging yet highly rewarding experience!

Tech Stack

TensorFlowKerasPythonGoogle ColabCNN

Key Highlights

  • Enhanced the dataset with robust data augmentation techniques to drastically improve model accuracy and robustness.
  • Built and trained a convolutional neural network (CNN) with optimized layers and dropout regularization.
  • Developed an automated pipeline to actively classify images and dynamically move them into respective category folders based on predictions.
  • Achieved high validation accuracy, ensuring reliable model performance on unseen test data.

Architecture Details

1. Data Preprocessing & Augmentation

  • Applying varied transformations explicitly to expand the dataset variation preventing overfitting and improving real-world generalization.

2. CNN Model Architecture & Automation

  • Built on TensorFlow & Keras, introducing Dropouts selectively to penalize excessive weight correlations yielding excellent validation accuracy.
  • Enforced an automated OS-level script sorting output files directly matching prediction classes into local directories.