vit-colon-cancer-classification

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

• Loss: 0.6794
• Accuracy: 0.8210

Model description

• Fine tuned vision transformer for classification of colon cancer.
• Four classes: Normal Tissue, Serrated Lesion, Adenoma, Adenocarcinoma

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: 2e-05
• train_batch_size: 10
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.8993	0.35	100	0.6462	0.7341
0.6042	0.71	200	0.6380	0.7432
0.6284	1.06	300	0.5628	0.7821
0.5494	1.42	400	0.5643	0.7788
0.5218	1.77	500	0.5478	0.7970
0.5053	2.13	600	0.5356	0.7846
0.4441	2.48	700	0.6928	0.7133
0.4492	2.84	800	0.4898	0.8078
0.429	3.19	900	0.5166	0.8020
0.3474	3.55	1000	0.5373	0.8061
0.337	3.9	1100	0.5442	0.7904
0.3243	4.26	1200	0.5171	0.8111
0.3003	4.61	1300	0.5463	0.8070
0.3127	4.96	1400	0.5122	0.8202
0.2587	5.32	1500	0.5807	0.8152
0.2434	5.67	1600	0.5392	0.8219
0.1996	6.03	1700	0.6343	0.8045
0.2033	6.38	1800	0.5855	0.8128
0.2056	6.74	1900	0.6516	0.8144
0.1927	7.09	2000	0.5770	0.8227
0.1688	7.45	2100	0.6153	0.8293
0.1566	7.8	2200	0.5994	0.8268
0.1406	8.16	2300	0.6192	0.8277
0.1381	8.51	2400	0.6334	0.8202
0.12	8.87	2500	0.6444	0.8136
0.104	9.22	2600	0.6709	0.8202
0.1049	9.57	2700	0.6752	0.8227
0.1349	9.93	2800	0.6980	0.8186
0.0846	10.28	2900	0.6794	0.8210

Framework versions

• Transformers 4.35.2
• Pytorch 2.0.1
• Datasets 2.15.0
• Tokenizers 0.15.0