resnet-18-feature-extraction

This model is a fine-tuned version of microsoft/resnet-18 on the imagefolder dataset. It achieves the following results on the evaluation set:

• Loss: 0.1485
• Accuracy: 0.95
• Precision: 0.9653
• Recall: 0.9789
• F1: 0.9720
• Roc Auc: 0.8505

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: 2e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1	Roc Auc
No log	0.8	2	0.6232	0.75	0.9636	0.7465	0.8413	0.7621
No log	1.8	4	0.6971	0.4875	1.0	0.4225	0.5941	0.7113
No log	2.8	6	0.7915	0.2875	1.0	0.1972	0.3294	0.5986
No log	3.8	8	0.8480	0.2875	1.0	0.1972	0.3294	0.5986
0.8651	4.8	10	0.9094	0.2562	1.0	0.1620	0.2788	0.5810
0.8651	5.8	12	0.7470	0.5625	1.0	0.5070	0.6729	0.7535
0.8651	6.8	14	0.5915	0.85	1.0	0.8310	0.9077	0.9155
0.8651	7.8	16	0.4817	0.8875	0.9844	0.8873	0.9333	0.8881
0.8651	8.8	18	0.3455	0.9187	0.9778	0.9296	0.9531	0.8815
0.5349	9.8	20	0.2966	0.9187	0.9708	0.9366	0.9534	0.8572
0.5349	10.8	22	0.2347	0.95	0.9653	0.9789	0.9720	0.8505
0.5349	11.8	24	0.2468	0.9313	0.9645	0.9577	0.9611	0.8400
0.5349	12.8	26	0.2310	0.9563	0.9720	0.9789	0.9754	0.8783
0.5349	13.8	28	0.2083	0.9313	0.9580	0.9648	0.9614	0.8157
0.3593	14.8	30	0.1840	0.9375	0.9521	0.9789	0.9653	0.7950
0.3593	15.8	32	0.1947	0.9375	0.9648	0.9648	0.9648	0.8435
0.3593	16.8	34	0.1837	0.9313	0.9517	0.9718	0.9617	0.7915
0.3593	17.8	36	0.1819	0.9437	0.9524	0.9859	0.9689	0.7985
0.3593	18.8	38	0.1924	0.9437	0.9650	0.9718	0.9684	0.8470
0.2737	19.8	40	0.1990	0.95	0.9653	0.9789	0.9720	0.8505
0.2737	20.8	42	0.1759	0.95	0.9718	0.9718	0.9718	0.8748
0.2737	21.8	44	0.1804	0.9313	0.9517	0.9718	0.9617	0.7915
0.2737	22.8	46	0.1666	0.9313	0.9517	0.9718	0.9617	0.7915
0.2737	23.8	48	0.1534	0.9437	0.9524	0.9859	0.9689	0.7985
0.2278	24.8	50	0.1612	0.9375	0.9521	0.9789	0.9653	0.7950
0.2278	25.8	52	0.1535	0.9437	0.9586	0.9789	0.9686	0.8228
0.2278	26.8	54	0.1568	0.9437	0.9716	0.9648	0.9682	0.8713
0.2278	27.8	56	0.2107	0.9375	0.9714	0.9577	0.9645	0.8678
0.2278	28.8	58	0.1592	0.9313	0.9517	0.9718	0.9617	0.7915
0.2057	29.8	60	0.1557	0.9375	0.9648	0.9648	0.9648	0.8435
0.2057	30.8	62	0.1714	0.9437	0.9650	0.9718	0.9684	0.8470
0.2057	31.8	64	0.1571	0.95	0.9653	0.9789	0.9720	0.8505
0.2057	32.8	66	0.1574	0.9375	0.9583	0.9718	0.9650	0.8192
0.2057	33.8	68	0.1423	0.9563	0.9720	0.9789	0.9754	0.8783
0.2	34.8	70	0.1677	0.9437	0.9650	0.9718	0.9684	0.8470
0.2	35.8	72	0.1560	0.9375	0.9583	0.9718	0.9650	0.8192
0.2	36.8	74	0.1594	0.9375	0.9521	0.9789	0.9653	0.7950
0.2	37.8	76	0.1512	0.9437	0.9586	0.9789	0.9686	0.8228
0.2	38.8	78	0.1396	0.9563	0.9655	0.9859	0.9756	0.8541
0.1838	39.8	80	0.1509	0.9375	0.9583	0.9718	0.9650	0.8192
0.1838	40.8	82	0.1529	0.95	0.9718	0.9718	0.9718	0.8748
0.1838	41.8	84	0.1506	0.95	0.9653	0.9789	0.9720	0.8505
0.1838	42.8	86	0.1549	0.95	0.9653	0.9789	0.9720	0.8505
0.1838	43.8	88	0.1331	0.9563	0.9655	0.9859	0.9756	0.8541
0.1872	44.8	90	0.1409	0.9437	0.9524	0.9859	0.9689	0.7985
0.1872	45.8	92	0.1639	0.9375	0.9583	0.9718	0.9650	0.8192
0.1872	46.8	94	0.1391	0.95	0.9589	0.9859	0.9722	0.8263
0.1872	47.8	96	0.1436	0.9563	0.9655	0.9859	0.9756	0.8541
0.1872	48.8	98	0.1442	0.9437	0.9586	0.9789	0.9686	0.8228
0.185	49.8	100	0.1485	0.95	0.9653	0.9789	0.9720	0.8505

Framework versions

• Transformers 4.24.0.dev0
• Pytorch 1.11.0+cu102
• Datasets 2.6.1
• Tokenizers 0.13.1