Dogs vs. Cats Computer Vision Classification

Project Goals

• Build a reliable deep-learning workflow for binary image classification.
• Compare model variants instead of relying on a single architecture pass.
• Practice the full pipeline from image resizing and normalization through evaluation and submission export.

Approach

The project uses image loading, rescaling, and generator-based training workflows to make larger image data manageable.

Multiple CNN configurations were explored, including different layer depths and training setups, so the notebook reads as an iteration trail rather than a single final run.

Result

One evaluated notebook run reports about 92.01% accuracy, making this a useful demonstration of practical CNN experimentation.

The value of the project is not only the score, but the process: preprocessing decisions, architecture changes, and evaluation formatting all appear in the workflow.

What It Shows

• Deep-learning fundamentals applied to image data.
• Comfort with TensorFlow/Keras workflows.
• Model iteration and benchmark-style project packaging.