segformer-b4-finetuned-segments-chargers-full-v4.1

This model is a fine-tuned version of nvidia/mit-b4 on the dskong07/chargers-full-v0.1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4050
• Mean Iou: 0.7302
• Mean Accuracy: 0.8405
• Overall Accuracy: 0.9155
• Accuracy Unlabeled: nan
• Accuracy Screen: 0.8874
• Accuracy Body: 0.9156
• Accuracy Cable: 0.6425
• Accuracy Plug: 0.8031
• Accuracy Void-background: 0.9539
• Iou Unlabeled: nan
• Iou Screen: 0.7807
• Iou Body: 0.8343
• Iou Cable: 0.5652
• Iou Plug: 0.5556
• Iou Void-background: 0.9154

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: 6e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Screen	Accuracy Body	Accuracy Cable	Accuracy Plug	Accuracy Void-background	Iou Unlabeled	Iou Screen	Iou Body	Iou Cable	Iou Plug	Iou Void-background
0.5049	2.2222	20	0.8068	0.5018	0.7473	0.8447	nan	0.7238	0.9311	0.4504	0.7647	0.8667	0.0	0.6849	0.7101	0.3604	0.4256	0.8298
0.2822	4.4444	40	0.5160	0.6209	0.7951	0.8640	nan	0.9440	0.7910	0.5499	0.7662	0.9246	nan	0.6072	0.7047	0.4638	0.4484	0.8802
0.1945	6.6667	60	0.4086	0.6900	0.8274	0.8978	nan	0.8896	0.8973	0.6177	0.7982	0.9342	nan	0.7376	0.8019	0.5163	0.4930	0.9009
0.2245	8.8889	80	0.4093	0.7007	0.8260	0.9020	nan	0.8361	0.9279	0.6242	0.8090	0.9328	nan	0.7540	0.8136	0.5345	0.4992	0.9021
0.1808	11.1111	100	0.3741	0.7021	0.8068	0.9051	nan	0.7754	0.9242	0.6154	0.7694	0.9498	nan	0.7222	0.8089	0.5384	0.5339	0.9070
0.1342	13.3333	120	0.3797	0.7106	0.8224	0.9066	nan	0.8693	0.9029	0.6329	0.7564	0.9504	nan	0.7614	0.8113	0.5539	0.5205	0.9061
0.1204	15.5556	140	0.4131	0.7112	0.8367	0.9063	nan	0.8737	0.9160	0.6331	0.8214	0.9392	nan	0.7638	0.8212	0.5585	0.5070	0.9057
0.1042	17.7778	160	0.3944	0.7180	0.8386	0.9096	nan	0.8884	0.9107	0.6271	0.8207	0.9461	nan	0.7674	0.8208	0.5567	0.5350	0.9103
0.093	20.0	180	0.3910	0.7231	0.8400	0.9121	nan	0.9020	0.9020	0.6366	0.8069	0.9526	nan	0.7731	0.8274	0.5605	0.5424	0.9119
0.0989	22.2222	200	0.3632	0.7260	0.8355	0.9142	nan	0.9002	0.9133	0.6173	0.7920	0.9547	nan	0.7818	0.8316	0.5560	0.5466	0.9139
0.0937	24.4444	220	0.3956	0.7241	0.8367	0.9130	nan	0.8941	0.9060	0.6165	0.8118	0.9550	nan	0.7743	0.8286	0.5584	0.5467	0.9128
0.0987	26.6667	240	0.4233	0.7240	0.8409	0.9125	nan	0.9040	0.9013	0.6209	0.8244	0.9537	nan	0.7752	0.8274	0.5574	0.5477	0.9124
0.0935	28.8889	260	0.4249	0.7252	0.8392	0.9124	nan	0.8810	0.9102	0.6346	0.8192	0.9512	nan	0.7748	0.8265	0.5619	0.5515	0.9115
0.0827	31.1111	280	0.4266	0.7241	0.8409	0.9124	nan	0.8800	0.9169	0.6387	0.8208	0.9480	nan	0.7736	0.8301	0.5659	0.5388	0.9121
0.0785	33.3333	300	0.4034	0.7285	0.8386	0.9144	nan	0.8847	0.9154	0.6273	0.8122	0.9533	nan	0.7804	0.8317	0.5631	0.5539	0.9135
0.0733	35.5556	320	0.4061	0.7316	0.8446	0.9150	nan	0.8993	0.9031	0.6489	0.8164	0.9554	nan	0.7853	0.8312	0.5658	0.5614	0.9144
0.0683	37.7778	340	0.4115	0.7266	0.8347	0.9148	nan	0.8581	0.9301	0.6239	0.8105	0.9511	nan	0.7685	0.8340	0.5604	0.5545	0.9154
0.1087	40.0	360	0.4317	0.7297	0.8459	0.9143	nan	0.9070	0.9005	0.6467	0.8211	0.9544	nan	0.7768	0.8286	0.5663	0.5620	0.9146
0.0634	42.2222	380	0.4252	0.7296	0.8426	0.9150	nan	0.8926	0.9141	0.6285	0.8250	0.9529	nan	0.7812	0.8341	0.5644	0.5541	0.9144
0.0786	44.4444	400	0.3969	0.7294	0.8408	0.9151	nan	0.8860	0.9139	0.6511	0.7993	0.9535	nan	0.7787	0.8333	0.5647	0.5551	0.9153
0.0693	46.6667	420	0.4133	0.7292	0.8422	0.9147	nan	0.8853	0.9138	0.6427	0.8168	0.9525	nan	0.7782	0.8334	0.5674	0.5523	0.9145
0.0663	48.8889	440	0.4050	0.7302	0.8405	0.9155	nan	0.8874	0.9156	0.6425	0.8031	0.9539	nan	0.7807	0.8343	0.5652	0.5556	0.9154

Framework versions

• Transformers 4.48.3
• Pytorch 2.6.0+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0