model1

This model is a fine-tuned version of nvidia/mit-b2 on the giuseppemartino/i-SAID_custom_or_1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1646
• Mean Iou: 0.2689
• Mean Accuracy: 0.3089
• Overall Accuracy: 0.3928
• Accuracy Background: nan
• Accuracy Ship: 0.7889
• Accuracy Small-vehicle: 0.3939
• Accuracy Tennis-court: 0.6399
• Accuracy Helicopter: nan
• Accuracy Basketball-court: 0.0
• Accuracy Ground-track-field: 0.4337
• Accuracy Swimming-pool: 0.6049
• Accuracy Harbor: 0.3386
• Accuracy Soccer-ball-field: 0.2551
• Accuracy Plane: 0.0001
• Accuracy Storage-tank: 0.0
• Accuracy Baseball-diamond: 0.5217
• Accuracy Large-vehicle: 0.3477
• Accuracy Bridge: 0.0
• Accuracy Roundabout: 0.0
• Iou Background: 0.0
• Iou Ship: 0.6137
• Iou Small-vehicle: 0.3354
• Iou Tennis-court: 0.6399
• Iou Helicopter: nan
• Iou Basketball-court: 0.0
• Iou Ground-track-field: 0.4084
• Iou Swimming-pool: 0.6049
• Iou Harbor: 0.3165
• Iou Soccer-ball-field: 0.2514
• Iou Plane: 0.0001
• Iou Storage-tank: 0.0
• Iou Baseball-diamond: 0.5217
• Iou Large-vehicle: 0.3418
• Iou Bridge: 0.0
• Iou Roundabout: 0.0

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: 8
• eval_batch_size: 8
• seed: 1337
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: polynomial
• training_steps: 840

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Ship	Accuracy Small-vehicle	Accuracy Tennis-court	Accuracy Helicopter	Accuracy Basketball-court	Accuracy Ground-track-field	Accuracy Swimming-pool	Accuracy Harbor	Accuracy Soccer-ball-field	Accuracy Plane	Accuracy Storage-tank	Accuracy Baseball-diamond	Accuracy Large-vehicle	Accuracy Bridge	Accuracy Roundabout	Iou Background	Iou Ship	Iou Small-vehicle	Iou Tennis-court	Iou Helicopter	Iou Basketball-court	Iou Ground-track-field	Iou Swimming-pool	Iou Harbor	Iou Soccer-ball-field	Iou Plane	Iou Storage-tank	Iou Baseball-diamond	Iou Large-vehicle	Iou Bridge	Iou Roundabout
1.1466	1.0	105	0.3419	0.0260	0.0279	0.0687	nan	0.0068	0.0036	0.3562	nan	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0240	0.0	0.0	0.0	0.0067	0.0036	0.3562	nan	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0240	0.0	0.0
0.3289	2.0	210	0.2301	0.1252	0.1441	0.2674	nan	0.5316	0.1793	0.6775	nan	0.0	0.0324	0.1854	0.1185	0.0	0.0	0.0	0.0	0.2923	0.0	0.0	0.0	0.4189	0.1612	0.6752	nan	0.0	0.0321	0.1854	0.1157	0.0	0.0	0.0	0.0	0.2898	0.0	0.0
0.1819	3.0	315	0.1965	0.1611	0.1937	0.3286	nan	0.7305	0.2842	0.4229	nan	0.0	0.3566	0.2424	0.1707	0.0739	0.0	0.0	0.0	0.4300	0.0	0.0	0.0	0.5605	0.2492	0.4229	nan	0.0	0.2817	0.2424	0.1637	0.0738	0.0	0.0	0.0	0.4223	0.0	0.0
0.1505	4.0	420	0.1760	0.1987	0.2352	0.3689	nan	0.7552	0.3079	0.5796	nan	0.0	0.4515	0.4367	0.2065	0.1437	0.0	0.0	0.0	0.4115	0.0	0.0	0.0	0.5715	0.2762	0.5790	nan	0.0	0.3752	0.4367	0.1957	0.1435	0.0	0.0	0.0	0.4029	0.0	0.0
0.1269	5.0	525	0.1688	0.2239	0.2616	0.3561	nan	0.8249	0.3133	0.5309	nan	0.0	0.3966	0.6398	0.2513	0.1975	0.0003	0.0	0.1336	0.3738	0.0	0.0	0.0	0.6006	0.2833	0.5309	nan	0.0	0.3711	0.6398	0.2378	0.1957	0.0003	0.0	0.1336	0.3661	0.0	0.0
0.1012	6.0	630	0.1763	0.2563	0.3036	0.3830	nan	0.7977	0.4801	0.6774	nan	0.0	0.4913	0.7772	0.2993	0.2702	0.0	0.0	0.2024	0.2541	0.0	0.0	0.0	0.6060	0.3488	0.6774	nan	0.0	0.4359	0.7767	0.2816	0.2638	0.0	0.0	0.2024	0.2515	0.0	0.0
0.0996	7.0	735	0.1687	0.2515	0.2906	0.3644	nan	0.7947	0.3775	0.5884	nan	0.0	0.4452	0.5756	0.2734	0.2140	0.0	0.0	0.4769	0.3225	0.0	0.0	0.0	0.6093	0.3246	0.5884	nan	0.0	0.4081	0.5756	0.2599	0.2128	0.0	0.0	0.4769	0.3174	0.0	0.0
0.0945	8.0	840	0.1646	0.2689	0.3089	0.3928	nan	0.7889	0.3939	0.6399	nan	0.0	0.4337	0.6049	0.3386	0.2551	0.0001	0.0	0.5217	0.3477	0.0	0.0	0.0	0.6137	0.3354	0.6399	nan	0.0	0.4084	0.6049	0.3165	0.2514	0.0001	0.0	0.5217	0.3418	0.0	0.0

Framework versions

• Transformers 4.35.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1