Modelpublic

sam1120/dropoff-utcustom-train-SF-RGB-b0_5

sourceHugging Faceotherupdated 3y agoView on Hugging Face

0likes12downloads

Model Card

dropoff-utcustom-train-SF-RGB-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: 0.2543
—Mean Iou: 0.6541
—Mean Accuracy: 0.6937
—Overall Accuracy: 0.9665
—Accuracy Unlabeled: nan
—Accuracy Dropoff: 0.3944
—Accuracy Undropoff: 0.9930
—Iou Unlabeled: nan
—Iou Dropoff: 0.3424
—Iou Undropoff: 0.9659

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: 5e-05
—trainbatchsize: 16
—evalbatchsize: 16
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_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
1.2123	3.33	10	1.1206	0.0793	0.1898	0.1888	nan	0.1908	0.1887	0.0	0.0494	0.1886
1.0927	6.67	20	1.0985	0.2196	0.5875	0.5351	nan	0.6450	0.5300	0.0	0.1290	0.5298
1.0578	10.0	30	0.9786	0.3662	0.7562	0.8622	nan	0.6400	0.8725	0.0	0.2367	0.8621
0.788	13.33	40	0.7940	0.4289	0.7505	0.9456	nan	0.5365	0.9646	0.0	0.3398	0.9468
0.6353	16.67	50	0.6206	0.4182	0.6840	0.9583	nan	0.3830	0.9850	0.0	0.2966	0.9581
0.6944	20.0	60	0.5213	0.4211	0.6766	0.9623	nan	0.3631	0.9901	0.0	0.3014	0.9620
0.5046	23.33	70	0.4765	0.4239	0.6796	0.9634	nan	0.3683	0.9910	0.0	0.3090	0.9628
0.4684	26.67	80	0.4643	0.3982	0.6347	0.9598	nan	0.2779	0.9914	0.0	0.2352	0.9593
0.4401	30.0	90	0.4483	0.4110	0.6507	0.9632	nan	0.3077	0.9936	0.0	0.2703	0.9627
0.4268	33.33	100	0.4366	0.6489	0.7001	0.9638	nan	0.4108	0.9895	nan	0.3347	0.9632
0.3939	36.67	110	0.4027	0.4272	0.6798	0.9650	nan	0.3670	0.9927	0.0	0.3171	0.9644
0.4472	40.0	120	0.4159	0.6428	0.6896	0.9638	nan	0.3887	0.9905	nan	0.3225	0.9632
0.3618	43.33	130	0.3765	0.6325	0.6671	0.9650	nan	0.3402	0.9939	nan	0.3006	0.9644
0.3456	46.67	140	0.3671	0.6395	0.6816	0.9643	nan	0.3715	0.9917	nan	0.3153	0.9637
0.3352	50.0	150	0.3572	0.6431	0.6839	0.9650	nan	0.3755	0.9923	nan	0.3218	0.9644
0.3143	53.33	160	0.3451	0.6351	0.6702	0.9651	nan	0.3467	0.9938	nan	0.3056	0.9646
0.3009	56.67	170	0.3357	0.6449	0.6941	0.9636	nan	0.3984	0.9898	nan	0.3267	0.9630
0.2765	60.0	180	0.3188	0.6458	0.6934	0.9641	nan	0.3965	0.9903	nan	0.3282	0.9634
0.2703	63.33	190	0.3179	0.6385	0.6732	0.9656	nan	0.3525	0.9940	nan	0.3119	0.9650
0.2746	66.67	200	0.3067	0.6385	0.6702	0.9662	nan	0.3456	0.9949	nan	0.3113	0.9656
0.2516	70.0	210	0.2992	0.6569	0.6968	0.9667	nan	0.4008	0.9929	nan	0.3477	0.9661
0.2503	73.33	220	0.2999	0.6671	0.7198	0.9659	nan	0.4497	0.9899	nan	0.3689	0.9652
0.2443	76.67	230	0.2816	0.6439	0.6750	0.9668	nan	0.3547	0.9952	nan	0.3215	0.9663
0.3757	80.0	240	0.2907	0.6593	0.7063	0.9659	nan	0.4215	0.9911	nan	0.3535	0.9652
0.2306	83.33	250	0.2767	0.6439	0.6807	0.9658	nan	0.3680	0.9935	nan	0.3226	0.9652
0.2216	86.67	260	0.2792	0.6583	0.7018	0.9663	nan	0.4115	0.9920	nan	0.3509	0.9657
0.3202	90.0	270	0.2681	0.6425	0.6789	0.9657	nan	0.3642	0.9936	nan	0.3199	0.9652
0.2174	93.33	280	0.2633	0.6467	0.6860	0.9657	nan	0.3791	0.9928	nan	0.3284	0.9651
0.2086	96.67	290	0.2658	0.6476	0.6900	0.9652	nan	0.3880	0.9920	nan	0.3306	0.9646
0.2042	100.0	300	0.2651	0.6486	0.6898	0.9655	nan	0.3873	0.9923	nan	0.3322	0.9649
0.2071	103.33	310	0.2597	0.6445	0.6792	0.9662	nan	0.3643	0.9941	nan	0.3233	0.9657
0.2097	106.67	320	0.2596	0.6615	0.7062	0.9665	nan	0.4206	0.9918	nan	0.3571	0.9658
0.3118	110.0	330	0.2557	0.6516	0.6928	0.9659	nan	0.3931	0.9924	nan	0.3380	0.9653
0.1956	113.33	340	0.2517	0.6494	0.6865	0.9664	nan	0.3794	0.9936	nan	0.3331	0.9658
0.201	116.67	350	0.2570	0.6573	0.7032	0.9658	nan	0.4151	0.9913	nan	0.3494	0.9651
0.1952	120.0	360	0.2543	0.6541	0.6937	0.9665	nan	0.3944	0.9930	nan	0.3424	0.9659

Framework versions

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