This model is a fine-tuned version of nvidia/mit-b3 on the dskong07/chargers-full-v0.1 dataset. It achieves the following results on the evaluation set:

Loss: 0.2675
Mean Iou: 0.8041
Mean Accuracy: 0.8712
Overall Accuracy: 0.9350
Accuracy Unlabeled: nan
Accuracy Screen: 0.7694
Accuracy Body: 0.9436
Accuracy Cable: 0.7715
Accuracy Plug: 0.9003
Accuracy Void-background: 0.9712
Iou Unlabeled: nan
Iou Screen: 0.7186
Iou Body: 0.8450
Iou Cable: 0.6753
Iou Plug: 0.8376
Iou Void-background: 0.9441

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
0.6506	2.2222	20	0.7818	0.5445	0.7593	0.8622	nan	0.6665	0.8659	0.5486	0.7951	0.9207	0.0	0.5753	0.7078	0.4054	0.7217	0.8569
0.435	4.4444	40	0.3972	0.7011	0.8078	0.8813	nan	0.8288	0.8134	0.6558	0.7876	0.9533	nan	0.6435	0.7211	0.5041	0.7552	0.8816
0.2948	6.6667	60	0.3063	0.7467	0.8376	0.9081	nan	0.7377	0.9184	0.7381	0.8463	0.9473	nan	0.6625	0.7904	0.5802	0.7875	0.9128
0.2005	8.8889	80	0.2800	0.7509	0.8233	0.9138	nan	0.6521	0.9449	0.7019	0.8601	0.9577	nan	0.6157	0.8022	0.6035	0.8092	0.9242
0.1762	11.1111	100	0.2633	0.7786	0.8579	0.9213	nan	0.7547	0.9300	0.7894	0.8576	0.9580	nan	0.6881	0.8148	0.6427	0.8201	0.9273
0.142	13.3333	120	0.2757	0.7790	0.8613	0.9217	nan	0.7705	0.9252	0.7758	0.8768	0.9582	nan	0.6801	0.8141	0.6413	0.8293	0.9303
0.1121	15.5556	140	0.2491	0.7916	0.8642	0.9293	nan	0.7772	0.9313	0.7556	0.8884	0.9686	nan	0.7121	0.8321	0.6505	0.8259	0.9373
0.1022	17.7778	160	0.2332	0.8041	0.8851	0.9326	nan	0.8106	0.9250	0.8261	0.8972	0.9667	nan	0.7340	0.8398	0.6668	0.8412	0.9387
0.1639	20.0	180	0.2549	0.7924	0.8591	0.9300	nan	0.7472	0.9360	0.7741	0.8655	0.9725	nan	0.7031	0.8342	0.6621	0.8247	0.9381
0.0953	22.2222	200	0.2447	0.8045	0.8806	0.9322	nan	0.7936	0.9322	0.8178	0.8939	0.9657	nan	0.7319	0.8379	0.6717	0.8433	0.9377
0.0899	24.4444	220	0.2581	0.7979	0.8640	0.9312	nan	0.7647	0.9434	0.7541	0.8905	0.9673	nan	0.7154	0.8346	0.6602	0.8404	0.9391
0.0826	26.6667	240	0.2403	0.8035	0.8692	0.9337	nan	0.7697	0.9426	0.7628	0.9010	0.9698	nan	0.7233	0.8424	0.6667	0.8451	0.9400
0.0765	28.8889	260	0.2622	0.8026	0.8739	0.9326	nan	0.7736	0.9396	0.7905	0.8987	0.9670	nan	0.7167	0.8392	0.6738	0.8432	0.9399
0.0812	31.1111	280	0.2668	0.8021	0.8738	0.9309	nan	0.7726	0.9236	0.8063	0.8948	0.9718	nan	0.7178	0.8341	0.6815	0.8415	0.9356
0.0804	33.3333	300	0.2394	0.8086	0.8767	0.9356	nan	0.7782	0.9371	0.8015	0.8937	0.9728	nan	0.7276	0.8458	0.6820	0.8444	0.9429
0.0682	35.5556	320	0.2538	0.8046	0.8799	0.9336	nan	0.8020	0.9316	0.8112	0.8861	0.9687	nan	0.7319	0.8416	0.6723	0.8363	0.9410
0.0768	37.7778	340	0.2576	0.8043	0.8712	0.9341	nan	0.7849	0.9357	0.7655	0.8973	0.9724	nan	0.7240	0.8417	0.6719	0.8421	0.9419
0.0789	40.0	360	0.2594	0.8040	0.8722	0.9346	nan	0.7789	0.9378	0.7740	0.8979	0.9723	nan	0.7208	0.8437	0.6760	0.8364	0.9432
0.0621	42.2222	380	0.2658	0.8036	0.8706	0.9343	nan	0.7709	0.9398	0.7780	0.8923	0.9721	nan	0.7170	0.8427	0.6752	0.8401	0.9432
0.063	44.4444	400	0.2609	0.8033	0.8686	0.9343	nan	0.7643	0.9386	0.7796	0.8864	0.9742	nan	0.7151	0.8433	0.6755	0.8397	0.9429
0.0625	46.6667	420	0.2655	0.8040	0.8725	0.9350	nan	0.7754	0.9391	0.7698	0.9057	0.9724	nan	0.7205	0.8455	0.6754	0.8352	0.9435
0.0808	48.8889	440	0.2675	0.8041	0.8712	0.9350	nan	0.7694	0.9436	0.7715	0.9003	0.9712	nan	0.7186	0.8450	0.6753	0.8376	0.9441

Framework versions

Transformers 4.48.3
Pytorch 2.6.0+cu124
Datasets 3.2.0
Tokenizers 0.21.0

JaesonGu/flare-segformer-mit-b3

segformer-b3-finetuned-segments-chargers-full-v3.1

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

JaesonGu/
flare-segformer-mit-b3