charliencku/mit-b0-finetuned-sidewalks

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

Train Loss: 0.6970
Validation Loss: 0.7208
Validation Mean Iou: 0.2419
Validation Mean Accuracy: 0.2928
Validation Overall Accuracy: 0.8003
Validation Accuracy Unlabeled: 0.0
Validation Accuracy Flat-road: 0.7802
Validation Accuracy Flat-sidewalk: 0.9508
Validation Accuracy Flat-crosswalk: 0.2997
Validation Accuracy Flat-cyclinglane: 0.7746
Validation Accuracy Flat-parkingdriveway: 0.2156
Validation Accuracy Flat-railtrack: 0.0
Validation Accuracy Flat-curb: 0.3495
Validation Accuracy Human-person: 0.5859
Validation Accuracy Human-rider: 0.0
Validation Accuracy Vehicle-car: 0.8950
Validation Accuracy Vehicle-truck: 0.0
Validation Accuracy Vehicle-bus: 0.0
Validation Accuracy Vehicle-tramtrain: 0.0
Validation Accuracy Vehicle-motorcycle: 0.0
Validation Accuracy Vehicle-bicycle: 0.5199
Validation Accuracy Vehicle-caravan: 0.0
Validation Accuracy Vehicle-cartrailer: 0.0
Validation Accuracy Construction-building: 0.8981
Validation Accuracy Construction-door: 0.0
Validation Accuracy Construction-wall: 0.4021
Validation Accuracy Construction-fenceguardrail: 0.2243
Validation Accuracy Construction-bridge: 0.0
Validation Accuracy Construction-tunnel: nan
Validation Accuracy Construction-stairs: 0.0
Validation Accuracy Object-pole: 0.2751
Validation Accuracy Object-trafficsign: 0.0
Validation Accuracy Object-trafficlight: 0.0
Validation Accuracy Nature-vegetation: 0.9210
Validation Accuracy Nature-terrain: 0.8262
Validation Accuracy Sky: 0.9260
Validation Accuracy Void-ground: 0.0
Validation Accuracy Void-dynamic: 0.0
Validation Accuracy Void-static: 0.1109
Validation Accuracy Void-unclear: 0.0
Validation Iou Unlabeled: 0.0
Validation Iou Flat-road: 0.6342
Validation Iou Flat-sidewalk: 0.8053
Validation Iou Flat-crosswalk: 0.2349
Validation Iou Flat-cyclinglane: 0.5463
Validation Iou Flat-parkingdriveway: 0.1949
Validation Iou Flat-railtrack: 0.0
Validation Iou Flat-curb: 0.2736
Validation Iou Human-person: 0.4927
Validation Iou Human-rider: 0.0
Validation Iou Vehicle-car: 0.7797
Validation Iou Vehicle-truck: 0.0
Validation Iou Vehicle-bus: 0.0
Validation Iou Vehicle-tramtrain: 0.0
Validation Iou Vehicle-motorcycle: 0.0
Validation Iou Vehicle-bicycle: 0.4065
Validation Iou Vehicle-caravan: 0.0
Validation Iou Vehicle-cartrailer: 0.0
Validation Iou Construction-building: 0.6731
Validation Iou Construction-door: 0.0
Validation Iou Construction-wall: 0.2720
Validation Iou Construction-fenceguardrail: 0.2128
Validation Iou Construction-bridge: 0.0
Validation Iou Construction-tunnel: nan
Validation Iou Construction-stairs: 0.0
Validation Iou Object-pole: 0.2210
Validation Iou Object-trafficsign: 0.0
Validation Iou Object-trafficlight: 0.0
Validation Iou Nature-vegetation: 0.8038
Validation Iou Nature-terrain: 0.6959
Validation Iou Sky: 0.8879
Validation Iou Void-ground: 0.0
Validation Iou Void-dynamic: 0.0
Validation Iou Void-static: 0.0890
Validation Iou Void-unclear: 0.0
Epoch: 2

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:

optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': 6e-05, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False}
training_precision: float32

Training results

Train Loss	Validation Loss	Validation Mean Iou	Validation Mean Accuracy	Validation Overall Accuracy	Validation Accuracy Flat-road	Validation Accuracy Flat-sidewalk	Validation Accuracy Flat-crosswalk	Validation Accuracy Flat-cyclinglane	Validation Accuracy Flat-parkingdriveway	Validation Accuracy Flat-curb	Validation Accuracy Human-person	Validation Accuracy Vehicle-car	Validation Accuracy Vehicle-bicycle	Validation Accuracy Construction-building	Validation Accuracy Construction-wall	Validation Accuracy Construction-fenceguardrail	Validation Accuracy Construction-tunnel	Validation Accuracy Object-pole	Validation Accuracy Nature-vegetation	Validation Accuracy Nature-terrain	Validation Accuracy Sky	Validation Accuracy Void-static	Validation Iou Flat-road	Validation Iou Flat-sidewalk	Validation Iou Flat-crosswalk	Validation Iou Flat-cyclinglane	Validation Iou Flat-parkingdriveway	Validation Iou Flat-curb	Validation Iou Human-person	Validation Iou Vehicle-car	Validation Iou Vehicle-bicycle	Validation Iou Construction-building	Validation Iou Construction-wall	Validation Iou Construction-fenceguardrail	Validation Iou Construction-tunnel	Validation Iou Object-pole	Validation Iou Nature-vegetation	Validation Iou Nature-terrain	Validation Iou Sky	Validation Iou Void-static	Epoch
1.4310	0.9029	0.1675	0.2127	0.7549	0.7962	0.9482	0.0470	0.4923	0.1047	0.1139	0.0073	0.9389	0.0002	0.7966	0.3058	0.0619	nan	0.0265	0.9220	0.7834	0.8833	0.0033	0.6033	0.7665	0.0470	0.3853	0.0964	0.1001	0.0073	0.5833	0.0002	0.5991	0.2315	0.0612	nan	0.0260	0.7446	0.6087	0.8297	0.0033	0
0.8498	0.7762	0.2121	0.2627	0.7858	0.8313	0.9411	0.1572	0.7171	0.1614	0.2595	0.4292	0.9269	0.1429	0.8122	0.4926	0.1613	nan	0.1157	0.9200	0.8737	0.9387	0.0501	0.6311	0.7981	0.1514	0.5360	0.1470	0.2161	0.3875	0.7166	0.1408	0.6622	0.2724	0.1557	nan	0.1054	0.7675	0.6138	0.8640	0.0456	1
0.6970	0.7208	0.2419	0.2928	0.8003	0.7802	0.9508	0.2997	0.7746	0.2156	0.3495	0.5859	0.8950	0.5199	0.8981	0.4021	0.2243	nan	0.2751	0.9210	0.8262	0.9260	0.1109	0.6342	0.8053	0.2349	0.5463	0.1949	0.2736	0.4927	0.7797	0.4065	0.6731	0.2720	0.2128	nan	0.2210	0.8038	0.6959	0.8879	0.0890	2

Framework versions

Transformers 4.29.2
TensorFlow 2.12.0
Datasets 2.12.0
Tokenizers 0.13.3

charliencku/mit-b0-finetuned-sidewalks

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

NSDT 3DConvert

UnrealSynth

DreamTexture.js