File size: 3,400 Bytes
743b3d0
 
 
f5c0f4b
 
743b3d0
 
 
 
 
 
 
 
 
 
 
f5c0f4b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
743b3d0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
---
license: apache-2.0
tags:
- image-segmentation
- vision
- generated_from_trainer
model-index:
- name: segformer-finetuned-sidewalk-50-epochs
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# segformer-finetuned-sidewalk-50-epochs

This model is a fine-tuned version of [nvidia/mit-b0](https://huggingface.co/nvidia/mit-b0) on the segments/sidewalk-semantic dataset.
It achieves the following results on the evaluation set:
- Loss: 0.7690
- Mean Iou: 0.2117
- Mean Accuracy: 0.2615
- Overall Accuracy: 0.7840
- Accuracy Unlabeled: nan
- Accuracy Flat-road: 0.6563
- Accuracy Flat-sidewalk: 0.9295
- Accuracy Flat-crosswalk: 0.3228
- Accuracy Flat-cyclinglane: 0.7303
- Accuracy Flat-parkingdriveway: 0.2743
- Accuracy Flat-railtrack: nan
- Accuracy Flat-curb: 0.3172
- Accuracy Human-person: 0.0109
- Accuracy Human-rider: 0.0
- Accuracy Vehicle-car: 0.9388
- Accuracy Vehicle-truck: 0.0
- Accuracy Vehicle-bus: 0.0
- Accuracy Vehicle-tramtrain: 0.0
- Accuracy Vehicle-motorcycle: 0.0
- Accuracy Vehicle-bicycle: 0.0
- Accuracy Vehicle-caravan: 0.0
- Accuracy Vehicle-cartrailer: 0.0
- Accuracy Construction-building: 0.9232
- Accuracy Construction-door: 0.0
- Accuracy Construction-wall: 0.2146
- Accuracy Construction-fenceguardrail: 0.0943
- Accuracy Construction-bridge: 0.0
- Accuracy Construction-tunnel: nan
- Accuracy Construction-stairs: 0.0
- Accuracy Object-pole: 0.2119
- Accuracy Object-trafficsign: 0.0
- Accuracy Object-trafficlight: 0.0
- Accuracy Nature-vegetation: 0.9083
- Accuracy Nature-terrain: 0.7748
- Accuracy Sky: 0.8846
- Accuracy Void-ground: 0.0
- Accuracy Void-dynamic: 0.0
- Accuracy Void-static: 0.1757
- Accuracy Void-unclear: 0.0
- Iou Unlabeled: nan
- Iou Flat-road: 0.5342
- Iou Flat-sidewalk: 0.8036
- Iou Flat-crosswalk: 0.2661
- Iou Flat-cyclinglane: 0.6165
- Iou Flat-parkingdriveway: 0.2254
- Iou Flat-railtrack: nan
- Iou Flat-curb: 0.1921
- Iou Human-person: 0.0109
- Iou Human-rider: 0.0
- Iou Vehicle-car: 0.7190
- Iou Vehicle-truck: 0.0
- Iou Vehicle-bus: 0.0
- Iou Vehicle-tramtrain: 0.0
- Iou Vehicle-motorcycle: 0.0
- Iou Vehicle-bicycle: 0.0
- Iou Vehicle-caravan: 0.0
- Iou Vehicle-cartrailer: 0.0
- Iou Construction-building: 0.6152
- Iou Construction-door: 0.0
- Iou Construction-wall: 0.1796
- Iou Construction-fenceguardrail: 0.0924
- Iou Construction-bridge: 0.0
- Iou Construction-tunnel: nan
- Iou Construction-stairs: 0.0
- Iou Object-pole: 0.1438
- Iou Object-trafficsign: 0.0
- Iou Object-trafficlight: 0.0
- Iou Nature-vegetation: 0.7901
- Iou Nature-terrain: 0.6467
- Iou Sky: 0.8389
- Iou Void-ground: 0.0
- Iou Void-dynamic: 0.0
- Iou Void-static: 0.0992
- Iou Void-unclear: 0.0

## 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: 16
- eval_batch_size: 16
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 50.0

### Training results



### Framework versions

- Transformers 4.19.0.dev0
- Pytorch 1.11.0+cu113
- Datasets 2.0.0
- Tokenizers 0.11.6