dropoff-utcustom-train-SF-RGBD-b0_5

This model is a fine-tuned version of nvidia/mit-b0 on the sam1120/dropoff-utcustom-TRAIN dataset. It achieves the following results on the evaluation set:

• Loss: 0.2608
• Mean Iou: 0.6161
• Mean Accuracy: 0.6630
• Overall Accuracy: 0.9623
• Accuracy Unlabeled: nan
• Accuracy Dropoff: 0.3365
• Accuracy Undropoff: 0.9894
• Iou Unlabeled: nan
• Iou Dropoff: 0.2705
• Iou Undropoff: 0.9617

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 120

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Unlabeled	Accuracy Dropoff	Accuracy Undropoff	Iou Unlabeled	Iou Dropoff	Iou Undropoff
0.9263	5.0	10	1.0370	0.2869	0.7147	0.7632	nan	0.6618	0.7675	0.0	0.1042	0.7565
0.8069	10.0	20	0.8622	0.4857	0.5062	0.9589	nan	0.0125	0.9999	nan	0.0124	0.9589
0.6851	15.0	30	0.6490	0.4876	0.5081	0.9586	nan	0.0167	0.9995	nan	0.0165	0.9586
0.5882	20.0	40	0.4739	0.3253	0.5085	0.9586	nan	0.0177	0.9994	0.0	0.0174	0.9585
0.53	25.0	50	0.4153	0.3375	0.5274	0.9584	nan	0.0573	0.9975	0.0	0.0542	0.9583
0.5009	30.0	60	0.4275	0.3835	0.6488	0.9475	nan	0.3230	0.9746	0.0	0.2037	0.9468
0.4699	35.0	70	0.3819	0.4158	0.6985	0.9578	nan	0.4157	0.9813	0.0	0.2904	0.9570
0.3946	40.0	80	0.3563	0.6183	0.6844	0.9585	nan	0.3854	0.9834	nan	0.2787	0.9579
0.3788	45.0	90	0.3259	0.6292	0.7011	0.9593	nan	0.4196	0.9827	nan	0.2998	0.9585
0.3412	50.0	100	0.3392	0.6170	0.6933	0.9562	nan	0.4066	0.9801	nan	0.2785	0.9555
0.3326	55.0	110	0.3214	0.6279	0.6914	0.9606	nan	0.3977	0.9851	nan	0.2958	0.9600
0.2954	60.0	120	0.3119	0.6261	0.6847	0.9613	nan	0.3831	0.9864	nan	0.2915	0.9607
0.3006	65.0	130	0.2853	0.5900	0.6223	0.9625	nan	0.2513	0.9934	nan	0.2180	0.9621
0.2715	70.0	140	0.3021	0.6314	0.6903	0.9620	nan	0.3938	0.9867	nan	0.3014	0.9614
0.276	75.0	150	0.2950	0.6243	0.6783	0.9619	nan	0.3690	0.9877	nan	0.2873	0.9613
0.2622	80.0	160	0.2843	0.6134	0.6651	0.9608	nan	0.3426	0.9876	nan	0.2665	0.9602
0.2395	85.0	170	0.2752	0.6050	0.6495	0.9613	nan	0.3094	0.9895	nan	0.2493	0.9608
0.2597	90.0	180	0.2813	0.6296	0.6874	0.9620	nan	0.3879	0.9869	nan	0.2979	0.9614
0.2294	95.0	190	0.2747	0.6106	0.6575	0.9615	nan	0.3259	0.9890	nan	0.2602	0.9609
0.2303	100.0	200	0.2606	0.6040	0.6462	0.9616	nan	0.3023	0.9902	nan	0.2468	0.9611
0.2335	105.0	210	0.2606	0.6080	0.6515	0.9619	nan	0.3130	0.9901	nan	0.2547	0.9614
0.2322	110.0	220	0.2619	0.6167	0.6631	0.9624	nan	0.3366	0.9896	nan	0.2715	0.9619
0.2116	115.0	230	0.2618	0.6183	0.6660	0.9624	nan	0.3427	0.9893	nan	0.2747	0.9618
0.2099	120.0	240	0.2608	0.6161	0.6630	0.9623	nan	0.3365	0.9894	nan	0.2705	0.9617

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu117
• Datasets 2.13.1
• Tokenizers 0.13.3