metadata

license: other
tags:
  - vision
  - image-segmentation
  - generated_from_trainer
model-index:
  - name: segformer-b0-finetuned-busigt2
    results: []

segformer-b0-finetuned-busigt2

This model is a fine-tuned version of nvidia/mit-b1 on the kasumi222/busigt5 dataset. It achieves the following results on the evaluation set:

Loss: 0.1923
Mean Iou: 0.4456
Mean Accuracy: 0.6990
Overall Accuracy: 0.6980
Per Category Iou: [0.0, 0.6613256012770924, 0.6755795107848668]
Per Category Accuracy: [nan, 0.6930571879874813, 0.7049128240257888]

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: 0.00013
train_batch_size: 20
eval_batch_size: 20
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Per Category Iou	Per Category Accuracy
0.44	0.77	20	0.4848	0.2588	0.4242	0.4165	[0.0, 0.3504134718879466, 0.42602074549863]	[nan, 0.3783013333071708, 0.47012910920923223]
0.3746	1.54	40	0.3811	0.3111	0.5104	0.5161	[0.0, 0.4842565751294053, 0.4491183491823408]	[nan, 0.5445819066586871, 0.4761423846712137]
0.3319	2.31	60	0.3142	0.2399	0.3819	0.3648	[0.0, 0.2677680368240358, 0.45190264244757206]	[nan, 0.279724508717023, 0.4840686746895068]
0.2211	3.08	80	0.2530	0.2975	0.4750	0.4814	[0.0, 0.4706420141972068, 0.4218805055880307]	[nan, 0.5129361209493069, 0.4371271301147258]
0.2097	3.85	100	0.2433	0.2900	0.4480	0.4460	[0.0, 0.43160250466254296, 0.43845878928113535]	[nan, 0.43601714952858417, 0.4600634808054265]
0.1869	4.62	120	0.2363	0.3632	0.5993	0.6166	[0.0, 0.6101054609134765, 0.4795020991617725]	[nan, 0.7024776555770306, 0.4960379563233556]
0.1632	5.38	140	0.2386	0.4353	0.6988	0.7093	[0.0, 0.6991547342129519, 0.6066344865447587]	[nan, 0.7611584761643136, 0.6364771236719307]
0.1939	6.15	160	0.2166	0.3374	0.5337	0.5229	[0.0, 0.46597040423540376, 0.5460862797699997]	[nan, 0.4692625114052214, 0.5981270739467682]
0.2074	6.92	180	0.2209	0.3219	0.5524	0.5826	[0.0, 0.5973031615126874, 0.3684272471225944]	[nan, 0.7324595053322476, 0.37233573901062894]
0.1243	7.69	200	0.2214	0.3890	0.6490	0.6624	[0.0, 0.6202500066793128, 0.5468835150043967]	[nan, 0.728845938760093, 0.5691879104528569]
0.1079	8.46	220	0.2349	0.3889	0.6489	0.6659	[0.0, 0.6357100818960788, 0.5310789782806924]	[nan, 0.7502934453088975, 0.5474674308445788]
0.1355	9.23	240	0.1988	0.4417	0.6835	0.6772	[0.0, 0.6347203785731718, 0.6904229863937434]	[nan, 0.6460698342931706, 0.7208882026363698]
0.1258	10.0	260	0.1985	0.4181	0.6597	0.6552	[0.0, 0.5971057628808589, 0.6571229098113397]	[nan, 0.6326629842926801, 0.6867490977333476]
0.1098	10.77	280	0.1959	0.4578	0.7091	0.7047	[0.0, 0.6732406203240918, 0.7002946319094014]	[nan, 0.6828090692358256, 0.7353112530851077]
0.097	11.54	300	0.1968	0.4401	0.6784	0.6719	[0.0, 0.6352861327900514, 0.6850189737146176]	[nan, 0.6398854081842887, 0.7169096389721925]
0.0844	12.31	320	0.1959	0.4164	0.6610	0.6637	[0.0, 0.6274020120203341, 0.6218360931423603]	[nan, 0.677014333787907, 0.645067517189047]
0.1543	13.08	340	0.2004	0.4261	0.6663	0.6626	[0.0, 0.618565945734453, 0.6597620117886153]	[nan, 0.6440009026067676, 0.6886109003283072]
0.0871	13.85	360	0.1967	0.4270	0.6699	0.6664	[0.0, 0.619536460193106, 0.6615374800234315]	[nan, 0.6488875371589471, 0.6909044252641271]
0.1012	14.62	380	0.1923	0.4456	0.6990	0.6980	[0.0, 0.6613256012770924, 0.6755795107848668]	[nan, 0.6930571879874813, 0.7049128240257888]

Framework versions

Transformers 4.21.3
Pytorch 1.12.1+cu113
Datasets 2.4.0
Tokenizers 0.12.1