This model is a fine-tuned version of facebook/dinov2-base on ~40gb of h&m clothing images. It is meant to clastify garments into their own category like "upper garment", "lower_garment", "underpants" etc

It achieves the following results on the evaluation set:

Loss: 0.1260
Accuracy: 0.9590

The results are even better because the dataset has sometimes wrong labels or several labels work well on the same piece of clothing.

How to use

from transformers import pipeline

pipe = pipeline("image-classification", model="cubbk/dinov2-base-finetuned-clothes-big")
result = pipe("https://static.zara.net/assets/public/9cee/ace9/4ac84b608e84/df271252ff15/04391788800-e1/04391788800-e1.jpg?ts=1758018702215&w=1500")

result[0]

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 144
eval_batch_size: 144
seed: 42
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.3022	1.0	583	0.2322	0.9210
0.2158	2.0	1166	0.1539	0.9501
0.1767	3.0	1749	0.1260	0.9590

Framework versions

Transformers 4.55.3
Pytorch 2.8.0+cu128
Datasets 4.0.0
Tokenizers 0.21.4