Small part of actual dataset meant for testing purposes.

A subset of the full dataset, containing 28,000 images (with 5,600 images per class), is used for initial training, validation, and testing. The dataset is split in a 14:3:3 ratio (70% training, 15% validation, 15% testing). This reduced dataset is intended for rapid experimentation to evaluate and compare the impact of various factors, including:
Data Augmentation Techniques
Class Imbalance Handling
Pretrained Backbone Selection
Freezing vs Unfreezing Layers
Model architecture depth
Fine-tuning strategies
Number of epochs
Learning rate adjustments

The goal is to identify the best-performing configuration on this test subset before applying it to the complete dataset to develop a state-of-the-art NSFW classification model.