sayakpaul/glpn-nyu-finetuned-diode-221122-030603

158次阅读

Training Loss	Epoch	Step	Validation Loss	Mae	Rmse	Abs Rel	Log Mae	Log Rmse	Delta1	Delta2	Delta3
1.3722	1.0	72	0.9950	2.2271	2.3100	3.3726	0.5488	0.6025	0.0176	0.0929	0.1687
0.4936	2.0	144	0.4493	0.3961	0.5488	0.4918	0.1648	0.2234	0.3827	0.6877	0.8553
0.4087	3.0	216	0.3971	0.3434	0.4737	0.4250	0.1437	0.1980	0.4393	0.7431	0.8979
0.3696	4.0	288	0.3895	0.3355	0.4652	0.4334	0.1405	0.1968	0.4412	0.7566	0.9174
0.3621	5.0	360	0.3926	0.3432	0.4766	0.4198	0.1416	0.1967	0.4372	0.7591	0.9095
0.3104	6.0	432	0.3615	0.3104	0.4447	0.3630	0.1281	0.1813	0.5032	0.7923	0.9237
0.2835	7.0	504	0.3744	0.3140	0.4506	0.3429	0.1326	0.1824	0.4801	0.7837	0.9135
0.2308	8.0	576	0.3615	0.3013	0.4458	0.3459	0.1243	0.1803	0.5334	0.8139	0.9173
0.2433	9.0	648	0.3609	0.3042	0.4374	0.3570	0.1258	0.1798	0.5223	0.8053	0.9229
0.2097	10.0	720	0.3628	0.3066	0.4312	0.3792	0.1276	0.1815	0.4973	0.8075	0.9254
0.1761	11.0	792	0.3637	0.3111	0.4413	0.3726	0.1286	0.1817	0.4982	0.7979	0.9223
0.22	12.0	864	0.3584	0.3077	0.4619	0.3345	0.1258	0.1808	0.5360	0.8007	0.9119
0.2087	13.0	936	0.3614	0.3078	0.4502	0.3513	0.1265	0.1808	0.5211	0.8055	0.9177
0.2036	14.0	1008	0.3601	0.3086	0.4486	0.3530	0.1267	0.1801	0.5184	0.8023	0.9189
0.2123	15.0	1080	0.3597	0.3054	0.4481	0.3462	0.1256	0.1798	0.5278	0.8055	0.9191

glpn-nyu-finetuned-diode-221122-030603

This model is a fine-tuned version of vinvino02/glpn-nyu on the diode-subset dataset.
It achieves the following results on the evaluation set:

Loss: 0.3597

Mae: 0.3054

Rmse: 0.4481

Abs Rel: 0.3462

Log Mae: 0.1256

Log Rmse: 0.1798

Delta1: 0.5278

Delta2: 0.8055

Delta3: 0.9191

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

train_batch_size: 24

eval_batch_size: 48

seed: 2022

optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08

lr_scheduler_type: linear

lr_scheduler_warmup_ratio: 0.2

num_epochs: 15

mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Mae Rmse Abs Rel Log Mae Log Rmse Delta1 Delta2 Delta3

1.3722 1.0 72 0.9950 2.2271 2.3100 3.3726 0.5488 0.6025 0.0176 0.0929 0.1687

0.4936 2.0 144 0.4493 0.3961 0.5488 0.4918 0.1648 0.2234 0.3827 0.6877 0.8553

0.4087 3.0 216 0.3971 0.3434 0.4737 0.4250 0.1437 0.1980 0.4393 0.7431 0.8979

0.3696 4.0 288 0.3895 0.3355 0.4652 0.4334 0.1405 0.1968 0.4412 0.7566 0.9174

0.3621 5.0 360 0.3926 0.3432 0.4766 0.4198 0.1416 0.1967 0.4372 0.7591 0.9095

0.3104 6.0 432 0.3615 0.3104 0.4447 0.3630 0.1281 0.1813 0.5032 0.7923 0.9237

0.2835 7.0 504 0.3744 0.3140 0.4506 0.3429 0.1326 0.1824 0.4801 0.7837 0.9135

0.2308 8.0 576 0.3615 0.3013 0.4458 0.3459 0.1243 0.1803 0.5334 0.8139 0.9173

0.2433 9.0 648 0.3609 0.3042 0.4374 0.3570 0.1258 0.1798 0.5223 0.8053 0.9229

0.2097 10.0 720 0.3628 0.3066 0.4312 0.3792 0.1276 0.1815 0.4973 0.8075 0.9254

0.1761 11.0 792 0.3637 0.3111 0.4413 0.3726 0.1286 0.1817 0.4982 0.7979 0.9223

0.22 12.0 864 0.3584 0.3077 0.4619 0.3345 0.1258 0.1808 0.5360 0.8007 0.9119

0.2087 13.0 936 0.3614 0.3078 0.4502 0.3513 0.1265 0.1808 0.5211 0.8055 0.9177

0.2036 14.0 1008 0.3601 0.3086 0.4486 0.3530 0.1267 0.1801 0.5184 0.8023 0.9189

0.2123 15.0 1080 0.3597 0.3054 0.4481 0.3462 0.1256 0.1798 0.5278 0.8055 0.9191