vit-l-16-food-classifier

This model is a fine-tuned version of google/vit-large-patch16-384 on 30VNFOODS and Custom Vietnamese foods dataset crawled on Google and ShopeeFood. It achieves the following results on the evaluation set:

• Loss: 0.3214
• Accuracy: 0.9155

Model description

Fine-tuned ViT-L-16 in 3 epochs.

Intended uses & limitations

This model is used for our project at UET-VNU Campathon 2024.

Training and evaluation data

30VNFOODS Dataset and self-crawled dataset.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	293	0.4138	0.8999
0.8179	2.0	586	0.3482	0.9151
0.8179	3.0	879	0.3214	0.9155

Framework versions

• Transformers 4.42.3
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1