vit-Facial-Confidence

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

Loss: 0.2560
Accuracy: 0.8970

Model description

Facial Confidence is an image classification model which takes a black and white image of a persons headshot and classifies it as confident or unconfident.

Intended uses & limitations

The model is intended to help with behavioral analysis tasks. The model is limited to black and white images where the image is a zoomed in headshot of a person (For best output the input image should be as zoomed in on the subjects face as possible without cutting any aspects of their head)

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 4
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.6103	0.0557	100	0.5715	0.7310
0.554	0.1114	200	0.5337	0.7194
0.4275	0.1671	300	0.5142	0.7549
0.5831	0.2228	400	0.5570	0.7345
0.5804	0.2786	500	0.4909	0.7660
0.5652	0.3343	600	0.4956	0.7764
0.4513	0.3900	700	0.4294	0.7972
0.4217	0.4457	800	0.4619	0.7924
0.435	0.5014	900	0.4563	0.7901
0.3943	0.5571	1000	0.4324	0.7917
0.4136	0.6128	1100	0.4131	0.8110
0.3302	0.6685	1200	0.4516	0.8054
0.4945	0.7242	1300	0.4135	0.8164
0.3729	0.7799	1400	0.4010	0.8139
0.4865	0.8357	1500	0.4145	0.8174
0.4011	0.8914	1600	0.4098	0.8112
0.4287	0.9471	1700	0.3914	0.8181
0.3644	1.0028	1800	0.3948	0.8188
0.3768	1.0585	1900	0.4044	0.8266
0.383	1.1142	2000	0.4363	0.8064
0.4011	1.1699	2100	0.4424	0.8025
0.4079	1.2256	2200	0.4384	0.7853
0.2791	1.2813	2300	0.4491	0.8089
0.3159	1.3370	2400	0.3863	0.8274
0.4306	1.3928	2500	0.3944	0.8158
0.3386	1.4485	2600	0.3835	0.8305
0.395	1.5042	2700	0.3812	0.8261
0.3041	1.5599	2800	0.3736	0.8312
0.3365	1.6156	2900	0.4420	0.8097
0.3697	1.6713	3000	0.3808	0.8353
0.3661	1.7270	3100	0.4046	0.8084
0.3208	1.7827	3200	0.4042	0.8328
0.3511	1.8384	3300	0.4113	0.8192
0.3246	1.8942	3400	0.3611	0.8377
0.3616	1.9499	3500	0.4207	0.8231
0.2726	2.0056	3600	0.3650	0.8342
0.1879	2.0613	3700	0.4334	0.8359
0.2981	2.1170	3800	0.3657	0.8435
0.227	2.1727	3900	0.3948	0.8399
0.3184	2.2284	4000	0.4229	0.8377
0.2391	2.2841	4100	0.3824	0.8405
0.2019	2.3398	4200	0.4628	0.8345
0.1931	2.3955	4300	0.3848	0.8448
0.238	2.4513	4400	0.3948	0.8398
0.2633	2.5070	4500	0.3779	0.8440
0.1829	2.5627	4600	0.3901	0.8455
0.2286	2.6184	4700	0.3797	0.8481
0.2123	2.6741	4800	0.4203	0.8502
0.266	2.7298	4900	0.4073	0.8455
0.1768	2.7855	5000	0.3750	0.8498
0.1659	2.8412	5100	0.3906	0.8427
0.1644	2.8969	5200	0.3833	0.8466
0.241	2.9526	5300	0.4071	0.8476
0.16	3.0084	5400	0.3691	0.8530
0.0788	3.0641	5500	0.4656	0.8514
0.1244	3.1198	5600	0.4990	0.8484
0.1423	3.1755	5700	0.5219	0.8475
0.1279	3.2312	5800	0.5687	0.8515
0.0974	3.2869	5900	0.5386	0.8458
0.065	3.3426	6000	0.5215	0.8454
0.0497	3.3983	6100	0.5161	0.8483
0.1871	3.4540	6200	0.5148	0.8523
0.0891	3.5097	6300	0.4915	0.8527
0.1375	3.5655	6400	0.5067	0.8509
0.1333	3.6212	6500	0.5272	0.8532
0.2635	3.6769	6600	0.5170	0.8516
0.0375	3.7326	6700	0.5148	0.8534
0.1286	3.7883	6800	0.4945	0.8543
0.091	3.8440	6900	0.4948	0.8540
0.1088	3.8997	7000	0.4985	0.8532
0.0598	3.9554	7100	0.4969	0.8514

Framework versions

Transformers 4.44.2
Pytorch 2.5.0+cu121
Datasets 3.0.2
Tokenizers 0.19.1

march18
/

FacialConfidence