This model is a fine-tuned version of google/vit-base-patch16-224-in21k on the imagefolder dataset. It achieves the following results on the evaluation set:

Loss: 1.2401
Accuracy: 0.625
Precision: 0.6207
Recall: 0.625
F1: 0.6035

Model description

A slightly more accurate model compared to previous 1.1 version. More information needed

Intended uses & limitations

This model is fined tune solely for face emotion recognition.

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 32
eval_batch_size: 32
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
No log	1.0	20	1.9487	0.3312	0.3554	0.3312	0.2830
No log	2.0	40	1.6735	0.4437	0.4238	0.4437	0.4232
No log	3.0	60	1.5359	0.4813	0.3990	0.4813	0.4272
No log	4.0	80	1.4249	0.5	0.4178	0.5	0.4443
No log	5.0	100	1.3733	0.5062	0.4753	0.5062	0.4653
No log	6.0	120	1.3513	0.5188	0.5076	0.5188	0.4908
No log	7.0	140	1.2377	0.6125	0.6163	0.6125	0.5976
No log	8.0	160	1.2354	0.6062	0.6131	0.6062	0.5961
No log	9.0	180	1.2574	0.575	0.5847	0.575	0.5728
No log	10.0	200	1.2493	0.5813	0.5912	0.5813	0.5776
No log	11.0	220	1.1954	0.5813	0.5795	0.5813	0.5730
No log	12.0	240	1.2283	0.5625	0.5651	0.5625	0.5598
No log	13.0	260	1.1984	0.5625	0.5800	0.5625	0.5643
No log	14.0	280	1.2308	0.5437	0.5523	0.5437	0.5414
No log	15.0	300	1.1665	0.5938	0.6005	0.5938	0.5935

Framework versions

Transformers 4.41.2
Pytorch 2.3.0
Datasets 2.19.1
Tokenizers 0.19.1

JamesJayamuni/emotion_classification_v1.2

emotion_classification_v1.2