ultrasound_plane_quality-vit-base-patch16-224-in21k

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

• Loss: 0.7986
• Accuracy: 0.6878
• Precision Macro: 0.4300
• Recall Macro: 0.4192
• F1 Macro: 0.4138
• Sensitivity Good: 0.8987
• Sensitivity Moderate: 0.0789
• Sensitivity Bad: 0.28
• Specificity Good: 0.3333
• Specificity Moderate: 0.9344
• Specificity Bad: 0.9235

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision Macro	Recall Macro	F1 Macro	Sensitivity Good	Sensitivity Moderate	Sensitivity Bad	Specificity Good	Specificity Moderate	Specificity Bad
0.7491	1.0	97	0.7897	0.7149	0.2383	0.3333	0.2779	1.0	0.0	0.0	0.0	1.0	1.0
0.8198	2.0	194	0.7772	0.7149	0.2383	0.3333	0.2779	1.0	0.0	0.0	0.0	1.0	1.0
0.7936	3.0	291	0.7657	0.7149	0.2383	0.3333	0.2779	1.0	0.0	0.0	0.0	1.0	1.0
0.7881	4.0	388	0.7843	0.7149	0.2383	0.3333	0.2779	1.0	0.0	0.0	0.0	1.0	1.0
0.7402	5.0	485	0.7698	0.7195	0.4082	0.3521	0.3177	0.9905	0.0658	0.0	0.0556	0.9863	1.0
0.8278	6.0	582	0.7597	0.7104	0.4491	0.3402	0.2988	0.9873	0.0132	0.02	0.0476	0.9809	0.9974
0.7460	7.0	679	0.7345	0.7149	0.2890	0.3367	0.2877	0.9968	0.0132	0.0	0.0476	0.9836	1.0
0.6520	8.0	776	0.7796	0.7036	0.5336	0.3749	0.3680	0.9462	0.1184	0.06	0.1825	0.9290	0.9949
0.6931	9.0	873	0.7693	0.6923	0.4729	0.3985	0.4009	0.9051	0.2105	0.08	0.2778	0.8962	0.9821
0.4950	10.0	970	0.7777	0.7081	0.5065	0.3770	0.3699	0.9525	0.1184	0.06	0.1905	0.9344	0.9923
0.5758	11.0	1067	0.7998	0.6923	0.4295	0.3685	0.3617	0.9335	0.0921	0.08	0.1905	0.9290	0.9796
0.5309	12.0	1164	0.7492	0.7285	0.5664	0.4749	0.4967	0.9209	0.2237	0.28	0.3333	0.9508	0.9541
0.6394	13.0	1261	0.7851	0.7036	0.4421	0.3628	0.3459	0.9620	0.0263	0.1	0.1032	0.9672	0.9847
0.4421	14.0	1358	0.8546	0.6584	0.4479	0.4365	0.4401	0.8259	0.2237	0.26	0.3968	0.8907	0.9107
0.3832	15.0	1455	0.8272	0.6742	0.4551	0.4283	0.4363	0.8608	0.1842	0.24	0.3651	0.8907	0.9388
0.3875	16.0	1552	0.8478	0.6765	0.4995	0.4103	0.4237	0.8734	0.1974	0.16	0.3095	0.8689	0.9796
0.4443	17.0	1649	0.8860	0.6810	0.4764	0.4401	0.4462	0.8513	0.3289	0.14	0.4048	0.8607	0.9617

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2