fruits-classification

This model is a fine-tuned version of google/vit-base-patch16-224 on the augmented Schram03/fruits dataset. It achieves the following results on the evaluation set:

• Loss: 0.3039
• Accuracy: 0.9280

Model description

This model classifies the 33 different types of fruit in the Schram03/fruits dataset.

Intended uses & limitations

With further developement it could be possible to differentiat foul fruits or specific subtypes of fruit. This however would need an extensive Dataset. Currently the model is able to differentiat the trained 33 types of fruit. However, there are limitations especially when using similar fruits like peaches and nectarines.

Training and evaluation data

80 % Train, 20 % Evaluation

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0003
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.5816	1.0	1070	0.5238	0.8887
0.3761	2.0	2140	0.3568	0.9182
0.2934	3.0	3210	0.2852	0.9322
0.2719	4.0	4280	0.2681	0.9388
0.253	5.0	5350	0.2705	0.9388

Framework versions

• Transformers 4.52.3
• Pytorch 2.7.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.1