stool-condition-classification

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

• Loss: 0.4237
• Auroc: 0.9418
• Accuracy: 0.9417
• Sensitivity: 0.9091
• Specificty: 0.9661
• Ppv: 0.9524
• Npv: 0.9344
• F1: 0.9302
• Model Selection: 0.9215

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

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: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Auroc	Accuracy	Sensitivity	Specificty	Ppv	Npv	F1	Model Selection
0.5076	0.98	100	0.5361	0.8538	0.7731	0.5393	0.9801	0.96	0.7061	0.6906	0.5592
0.4086	1.96	200	0.4857	0.8728	0.7836	0.6011	0.9453	0.9068	0.7280	0.7230	0.6558
0.5208	2.94	300	0.5109	0.8059	0.7599	0.6124	0.8905	0.8321	0.7218	0.7055	0.7218
0.474	3.92	400	0.5212	0.8601	0.7995	0.6180	0.9602	0.9322	0.7395	0.7432	0.6578
0.4285	4.9	500	0.4511	0.8728	0.7757	0.7472	0.8010	0.7688	0.7816	0.7578	0.9462
0.3506	5.88	600	0.4716	0.8691	0.8047	0.6798	0.9154	0.8768	0.7635	0.7658	0.7644
0.4239	6.86	700	0.5043	0.8517	0.8100	0.6685	0.9353	0.9015	0.7611	0.7677	0.7332
0.2447	7.84	800	0.5804	0.8592	0.8074	0.6910	0.9104	0.8723	0.7689	0.7712	0.7806
0.1739	8.82	900	0.6225	0.8562	0.8074	0.7135	0.8905	0.8523	0.7783	0.7768	0.8229
0.2888	9.8	1000	0.5807	0.8570	0.8047	0.7528	0.8507	0.8171	0.7953	0.7836	0.9021

Framework versions

• Transformers 4.38.2
• Pytorch 2.0.1
• Datasets 2.14.7
• Tokenizers 0.15.2