segformer-b0-mars-testbed-fewshot

This model is a fine-tuned version of nvidia/mit-b0 on the Tani04/mars_testbed_terrain dataset. It achieves the following results on the evaluation set:

• Loss: 5.5421
• Mean Iou: 0.3524
• Mean Accuracy: 0.4470
• Overall Accuracy: 0.6053
• Accuracy Flat Bedrock: 0.0
• Accuracy Flat Gravel: 0.8431
• Accuracy Hard Gravel: 0.1552
• Accuracy Obstacle: 0.7895
• Iou Flat Bedrock: 0.0
• Iou Flat Gravel: 0.7604
• Iou Hard Gravel: 0.1121
• Iou Obstacle: 0.5371

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: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Use 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: 60

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Flat Bedrock	Accuracy Flat Gravel	Accuracy Hard Gravel	Accuracy Obstacle	Iou Flat Bedrock	Iou Flat Gravel	Iou Hard Gravel	Iou Obstacle
0.9524	2.5	20	2.7762	0.2467	0.4256	0.5007	0.0	0.9625	0.1759	0.5642	0.0	0.4426	0.1131	0.4309
1.9458	5.0	40	3.0021	0.3148	0.4469	0.5622	0.0	0.9322	0.1733	0.6820	0.0	0.6568	0.1131	0.4893
0.5475	7.5	60	3.3026	0.3227	0.4611	0.5852	0.0	0.9585	0.1652	0.7207	0.0	0.6585	0.1147	0.5174
0.3696	10.0	80	3.7134	0.3445	0.4580	0.6020	0.0	0.9118	0.1511	0.7693	0.0	0.7352	0.1082	0.5347
0.3826	12.5	100	4.1692	0.3456	0.4555	0.5962	0.0	0.9079	0.1575	0.7565	0.0	0.7465	0.1107	0.5252
0.2651	15.0	120	4.1228	0.3453	0.4423	0.5933	0.0	0.8489	0.1522	0.7683	0.0	0.7503	0.1071	0.5237
0.3288	17.5	140	4.2263	0.3515	0.4577	0.6159	0.0	0.8787	0.1492	0.8030	0.0	0.7429	0.1106	0.5527
0.1629	20.0	160	4.6687	0.3453	0.4441	0.5827	0.0	0.8782	0.1597	0.7385	0.0	0.7653	0.1086	0.5073
0.377	22.5	180	4.5768	0.3475	0.4440	0.5953	0.0	0.8478	0.1601	0.7680	0.0	0.7527	0.1124	0.5250
0.3951	25.0	200	4.6531	0.3393	0.4327	0.5730	0.0	0.8427	0.1562	0.7317	0.0	0.7553	0.1053	0.4965
0.1941	27.5	220	4.5228	0.3530	0.4543	0.6116	0.0	0.8698	0.1508	0.7966	0.0	0.7556	0.1104	0.5462
0.237	30.0	240	4.5266	0.3526	0.4541	0.6031	0.0	0.8839	0.1580	0.7744	0.0	0.7650	0.1128	0.5325
0.2397	32.5	260	5.0232	0.3486	0.4460	0.5950	0.0	0.8608	0.1578	0.7656	0.0	0.7606	0.1111	0.5229
0.1909	35.0	280	5.4829	0.3564	0.4563	0.6200	0.0	0.8635	0.1462	0.8155	0.0	0.7600	0.1102	0.5554
0.1391	37.5	300	5.3365	0.3553	0.4581	0.6153	0.0	0.8811	0.1515	0.8000	0.0	0.7610	0.1123	0.5481
0.2368	40.0	320	5.4860	0.3527	0.4525	0.6051	0.0	0.8730	0.1551	0.7820	0.0	0.7635	0.1116	0.5357
0.1269	42.5	340	5.1693	0.3508	0.4462	0.5977	0.0	0.8546	0.1590	0.7712	0.0	0.7646	0.1124	0.5263
0.0917	45.0	360	5.3999	0.3552	0.4541	0.6138	0.0	0.8633	0.1517	0.8015	0.0	0.7618	0.1119	0.5471
2.1454	47.5	380	6.0907	0.3528	0.4518	0.6062	0.0	0.8685	0.1520	0.7867	0.0	0.7640	0.1102	0.5371
0.1348	50.0	400	5.6524	0.3541	0.4487	0.6087	0.0	0.8456	0.1535	0.7958	0.0	0.7633	0.1119	0.5411
0.1323	52.5	420	5.7016	0.3532	0.4492	0.6062	0.0	0.8531	0.1549	0.7888	0.0	0.7632	0.1120	0.5376
0.1165	55.0	440	5.6639	0.3502	0.4419	0.6033	0.0	0.8218	0.1545	0.7913	0.0	0.7538	0.1115	0.5355
0.1574	57.5	460	5.8704	0.3532	0.4474	0.6085	0.0	0.8378	0.1554	0.7963	0.0	0.7585	0.1132	0.5411
0.0974	60.0	480	5.5421	0.3524	0.4470	0.6053	0.0	0.8431	0.1552	0.7895	0.0	0.7604	0.1121	0.5371

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3