RoadSense高清街道分割開源模型 - 輕量級免費部署，精準分割人行道場景

首頁

Roadsense High Definition Street Segmentation

由iammartian0開發

基於SegFormer架構的輕量級圖像分割模型，專門針對人行道場景進行微調

圖像分割

Transformers

開源協議:其他 #人行道分割 #輕量級模型 #城市街景

下載量 63

發布時間 : 7/7/2023

模型概述

該模型是基於MIT-B0架構的SegFormer模型，在segments/sidewalk-semantic數據集上微調，用於人行道場景的語義分割任務。

模型特點

輕量級設計

基於B0級別的SegFormer架構，適合資源受限環境部署

人行道場景優化

專門針對人行道相關場景進行微調，提升相關類別識別準確率

多類別識別

可識別道路、人行道、建築物、車輛等40+類別

模型能力

語義分割

場景理解

道路元素識別

城市景觀分析

使用案例

智慧城市

人行道維護監測

自動檢測人行道損壞或異常區域

平坦人行道識別準確率達96.11%

交通基礎設施分析

識別道路標誌、交通燈等基礎設施

自動駕駛

道路場景理解

為自動駕駛系統提供環境感知能力

車輛汽車識別準確率達93.32%

🚀 segformer-b0-finetuned-segments-sidewalk

該模型是 nvidia/mit-b0 在 None 數據集上的微調版本，可用於圖像分割任務。在評估集上，它取得了一系列的評估指標結果，為圖像分割任務提供了有效的解決方案。

🚀 快速開始

本模型可藉助 transformers 庫使用，以下是使用示例代碼（此處暫未給出原文檔示例，你可根據實際情況補充）：

# 示例代碼
from transformers import AutoModelForImageSegmentation, AutoImageProcessor
import torch
from PIL import Image
import requests

model = AutoModelForImageSegmentation.from_pretrained("segformer-b0-finetuned-segments-sidewalk")
image_processor = AutoImageProcessor.from_pretrained("segformer-b0-finetuned-segments-sidewalk")

url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/segmentation_example.png"
image = Image.open(requests.get(url, stream=True).raw)

inputs = image_processor(images=image, return_tensors="pt")

with torch.no_grad():
    outputs = model(**inputs)

# 後續處理輸出結果

📚 詳細文檔

模型描述

此模型是 nvidia/mit-b0 在 None 數據集上的微調版本。它在圖像分割任務的評估集上取得了如下結果：

評估指標	值
Loss	0.5449
Mean Iou	0.3292
Mean Accuracy	0.3907
Overall Accuracy	0.8555
Accuracy Unlabeled	nan
Accuracy Flat-road	0.8585
Accuracy Flat-sidewalk	0.9611
Accuracy Flat-crosswalk	0.7673
Accuracy Flat-cyclinglane	0.8223
Accuracy Flat-parkingdriveway	0.5127
Accuracy Flat-railtrack	nan
Accuracy Flat-curb	0.4937
Accuracy Human-person	0.7164
Accuracy Human-rider	0.0
Accuracy Vehicle-car	0.9332
Accuracy Vehicle-truck	0.0
Accuracy Vehicle-bus	nan
Accuracy Vehicle-tramtrain	nan
Accuracy Vehicle-motorcycle	0.0
Accuracy Vehicle-bicycle	0.3858
Accuracy Vehicle-caravan	0.0
Accuracy Vehicle-cartrailer	0.0
Accuracy Construction-building	0.9040
Accuracy Construction-door	0.0
Accuracy Construction-wall	0.5848
Accuracy Construction-fenceguardrail	0.4417
Accuracy Construction-bridge	0.0
Accuracy Construction-tunnel	nan
Accuracy Construction-stairs	0.0
Accuracy Object-pole	0.3156
Accuracy Object-trafficsign	0.0
Accuracy Object-trafficlight	0.0
Accuracy Nature-vegetation	0.9413
Accuracy Nature-terrain	0.8456
Accuracy Sky	0.9600
Accuracy Void-ground	0.0
Accuracy Void-dynamic	0.0
Accuracy Void-static	0.2780
Accuracy Void-unclear	0.0
Iou Unlabeled	nan
Iou Flat-road	0.7447
Iou Flat-sidewalk	0.8755
Iou Flat-crosswalk	0.6244
Iou Flat-cyclinglane	0.7325
Iou Flat-parkingdriveway	0.3997
Iou Flat-railtrack	nan
Iou Flat-curb	0.3974
Iou Human-person	0.4985
Iou Human-rider	0.0
Iou Vehicle-car	0.7798
Iou Vehicle-truck	0.0
Iou Vehicle-bus	nan
Iou Vehicle-tramtrain	nan
Iou Vehicle-motorcycle	0.0
Iou Vehicle-bicycle	0.2904
Iou Vehicle-caravan	0.0
Iou Vehicle-cartrailer	0.0
Iou Construction-building	0.7233
Iou Construction-door	0.0
Iou Construction-wall	0.4555
Iou Construction-fenceguardrail	0.3734
Iou Construction-bridge	0.0
Iou Construction-tunnel	nan
Iou Construction-stairs	0.0
Iou Object-pole	0.2484
Iou Object-trafficsign	0.0
Iou Object-trafficlight	0.0
Iou Nature-vegetation	0.8451
Iou Nature-terrain	0.7346
Iou Sky	0.9161
Iou Void-ground	0.0
Iou Void-dynamic	0.0
Iou Void-static	0.2359
Iou Void-unclear	0.0

訓練和評估數據

更多信息待補充。

訓練過程

訓練超參數

訓練過程中使用了以下超參數：

超參數	值
learning_rate	6e - 05
train_batch_size	8
eval_batch_size	8
seed	42
optimizer	Adam with betas=(0.9, 0.999) and epsilon=1e - 08
lr_scheduler_type	linear
num_epochs	20

訓練結果

| 訓練損失 | 輪數 | 步數 | 驗證損失 | 平均交併比 | 平均準確率 | 總體準確率 | 未標記準確率 | 平坦道路準確率 | 平坦人行道準確率 | 平坦人行橫道準確率 | 平坦自行車道準確率 | 平坦停車場車道準確率 | 平坦鐵軌準確率 | 平坦路緣準確率 | 人類 - 行人準確率 | 人類 - 騎手準確率 | 車輛 - 汽車準確率 | 車輛 - 卡車準確率 | 車輛 - 公交車準確率 | 車輛 - 電車/火車準確率 | 車輛 - 摩托車準確率 | 車輛 - 自行車準確率 | 車輛 - 大篷車準確率 | 車輛 - 汽車拖車準確率 | 建築 - 建築物準確率 | 建築 - 門準確率 | 建築 - 牆準確率 | 建築 - 圍欄/護欄準確率 | 建築 - 橋樑準確率 | 建築 - 隧道準確率 | 建築 - 樓梯準確率 | 物體 - 杆子準確率 | 物體 - 交通標誌準確率 | 物體 - 交通燈準確率 | 自然 - 植被準確率 | 自然 - 地形準確率 | 天空準確率 | 空白 - 地面準確率 | 空白 - 動態準確率 | 空白 - 靜態準確率 | 空白 - 不明確準確率 | 未標記交併比 | 平坦道路交併比 | 平坦人行道交併比 | 平坦人行橫道交併比 | 平坦自行車道交併比 | 平坦停車場車道交併比 | 平坦鐵軌交併比 | 平坦路緣交併比 | 人類 - 行人交併比 | 人類 - 騎手交併比 | 車輛 - 汽車交併比 | 車輛 - 卡車交併比 | 車輛 - 公交車交併比 | 車輛 - 電車/火車交併比 | 車輛 - 摩托車交併比 | 車輛 - 自行車交併比 | 車輛 - 大篷車交併比 | 車輛 - 汽車拖車交併比 | 建築 - 建築物交併比 | 建築 - 門交併比 | 建築 - 牆交併比 | 建築 - 圍欄/護欄交併比 | 建築 - 橋樑交併比 | 建築 - 隧道交併比 | 建築 - 樓梯交併比 | 物體 - 杆子交併比 | 物體 - 交通標誌交併比 | 物體 - 交通燈交併比 | 自然 - 植被交併比 | 自然 - 地形交併比 | 天空交併比 | 空白 - 地面交併比 | 空白 - 動態交併比 | 空白 - 靜態交併比 | 空白 - 不明確交併比 | | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | ---- | | 1.4172 | 1.87 | 200 | 1.2183 | 0.1696 | 0.2214 | 0.7509 | nan | 0.8882 | 0.9199 | 0.0 | 0.4200 | 0.0164 | nan | 0.0 | 0.0 | 0.0 | 0.8778 | 0.0 | nan | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.8448 | 0.0 | 0.0 | 0.0 | 0.0 | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.9430 | 0.8044 | 0.9274 | 0.0 | 0.0 | 0.0 | 0.0 | nan | 0.5435 | 0.8135 | 0.0 | 0.3743 | 0.0160 | nan | 0.0 | 0.0 | 0.0 | 0.6044 | 0.0 | nan | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.5373 | 0.0 | 0.0 | 0.0 | 0.0 | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.7516 | 0.6550 | 0.7928 | 0.0 | 0.0 | 0.0 | 0.0 | | 1.1152 | 3.74 | 400 | 0.8946 | 0.1947 | 0.2441 | 0.7852 | nan | 0.8535 | 0.9471 | 0.0 | 0.7379 | 0.2453 | nan | 0.0398 | 0.0 | 0.0 | 0.8882 | 0.0 | nan | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.8746 | 0.0 | 0.0061 | 0.0 | 0.0 | nan | 0.0 | 0.0014 | 0.0 | 0.0 | 0.9526 | 0.8285 | 0.9448 | 0.0 | 0.0 | 0.0019 | 0.0 | nan | 0.6355 | 0.8321 | 0.0 | 0.5529 | 0.1940 | nan | 0.0392 | 0.0 | 0.0 | 0.6807 | 0.0 | nan | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.5913 | 0.0 | 0.0061 | 0.0 | 0.0 | nan | 0.0 | 0.0014 | 0.0 | 0.0 | 0.7701 | 0.6777 | 0.8567 | 0.0 | 0.0 | 0.0019 | 0.0 | | 0.6637 | 5.61 | 600 | 0.7447 | 0.2349 | 0.2841 | 0.8104 | nan | 0.8589 | 0.9451 | 0.4455 | 0.8008 | 0.3753 | nan | 0.3267 | 0.0380 | 0.0 | 0.8920 | 0.0 | nan | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.9227 | 0.0 | 0.0938 | 0.0 | 0.0 | nan | 0.0 | 0.0167 | 0.0 | 0.0 | 0.9291 | 0.8677 | 0.9557 | 0.0 | 0.0 | 0.0562 | 0.0 | nan | 0.6768 | 0.8543 | 0.4064 | 0.6414 | 0.2914 | nan | 0.2749 | 0.0376 | 0.0 | 0.7268 | 0.0 | nan | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.6078 | 0.0 | 0.0879 | 0.0 | 0.0 | nan | 0.0 | 0.0164 | 0.0 | 0.0 | 0.8005 | 0.6817 | 0.8918 | 0.0 | 0.0 | 0.0525 | 0.0 | | 0.673 | 7.48 | 800 | 0.6631 | 0.2691 | 0.3202 | 0.8278 | nan | 0.8387 | 0.9575 | 0.6176 | 0.7938 | 0.4208 | nan | 0.3575 | 0.3977 | 0.0 | 0.9264 | 0.0 | nan | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.9068 | 0.0 | 0.4035 | 0.0 | 0.0 | nan | 0.0 | 0.1137 | 0.0 | 0.0 | 0.9495 | 0.8165 | 0.9453 | 0.0 | 0.0 | 0.1599 | 0.0 | nan | 0.7042 | 0.8567 | 0.5239 | 0.6600 | 0.3246 | nan | 0.3003 | 0.3212 | 0.0 | 0.7246 | 0.0 | nan | nan | 0.0 | 0.0 | 0.0 | 0.0 | 0.6749 | 0.0 | 0.3113 | 0.0 | 0.0 | nan | 0.0 | 0.1038 | 0.0 | 0.0 | 0.8147 | 0.7070 | 0.9008 | 0.0 | 0.0 | 0.1445 | 0.0 | | 0.502 | 9.35 | 1000 | 0.6249 | 0.2818 | 0.3371 | 0.8345 | nan | 0.8332 | 0.9538 | 0.7158 | 0.8344 | 0.4079 | nan | 0.4420 | 0.4941 | 0.0 | 0.9275 | 0.0 | nan | nan | 0.0 | 0.0172 | 0.0 | 0.0 | 0.9102 | 0.0 | 0.4787 | 0.0253 | 0.0 | nan | 0.0 | 0.1454 | 0.0 | 0.0 | 0.9460 | 0.8350 | 0.9588 | 0.0 | 0.0 | 0.1887 | 0.0 | nan | 0.7176 | 0.8635 | 0.6035 | 0.6519 | 0.3246 | nan | 0.3545 | 0.3720 | 0.0 | 0.7524 | 0.0 | nan | nan | 0.0 | 0.0172 | 0.0 | 0.0 | 0.6861 | 0.0 | 0.3286 | 0.0250 | 0.0 | nan | 0.0 | 0.1309 | 0.0 | 0.0 | 0.8335 | 0.7300 | 0.9037 | 0.0 | 0.0 | 0.1584 | 0.0 | | 0.9687 | 11.21 | 1200 | 0.5786 | 0.3093 | 0.3675 | 0.8471 | nan | 0.8703 | 0.9504 | 0.7382 | 0.7705 | 0.5297 | nan | 0.4804 | 0.6250 | 0.0 | 0.9168 | 0.0 | nan | nan | 0.0 | 0.1397 | 0.0 | 0.0 | 0.9228 | 0.0 | 0.5710 | 0.3183 | 0.0 | nan | 0.0 | 0.2252 | 0.0 | 0.0 | 0.9314 | 0.8840 | 0.9536 | 0.0 | 0.0 | 0.1981 | 0.0 | nan | 0.7380 | 0.8743 | 0.5825 | 0.7093 | 0.3829 | nan | 0.3743 | 0.4600 | 0.0 | 0.7727 | 0.0 | nan | nan | 0.0 | 0.1372 | 0.0 | 0.0 | 0.7008 | 0.0 | 0.4315 | 0.2847 | 0.0 | nan | 0.0 | 0.1930 | 0.0 | 0.0 | 0.8397 | 0.7121 | 0.9109 | 0.0 | 0.0 | 0.1761 | 0.0 | | 0.4681 | 13.08 | 1400 | 0.5759 | 0.3106 | 0.3665 | 0.8462 | nan | 0.8586 | 0.9572 | 0.5158 | 0.8121 | 0.5195 | nan | 0.4539 | 0.6944 | 0.0 | 0.9308 | 0.0 | nan | nan | 0.0 | 0.2759 | 0.0 | 0.0 | 0.9126 | 0.0 | 0.4927 | 0.3145 | 0.0 | nan | 0.0 | 0.2566 | 0.0 | 0.0 | 0.9396 | 0.8736 | 0.9644 | 0.0 | 0.0 | 0.2226 | 0.0 | nan | 0.7134 | 0.8742 | 0.5009 | 0.7146 | 0.4018 | nan | 0.3726 | 0.4661 | 0.0 | 0.7674 | 0.0 | nan | nan | 0.0 | 0.2501 | 0.0 | 0.0 | 0.6997 | 0.0 | 0.3933 | 0.2827 | 0.0 | nan | 0.0 | 0.2137 | 0.0 | 0.0 | 0.8377 | 0.7212 | 0.9109 | 0.0 | 0.0 | 0.1964 | 0.0 | | 0.5374 | 14.95 | 1600 | 0.5534 | 0.3232 | 0.3823 | 0.8518 | nan | 0.8607 | 0.9545 | 0.7138 | 0.8398 | 0.5129 | nan | 0.4823 | 0.7055 | 0.0 | 0.9225 | 0.0 | nan | nan | 0.0 | 0.3058 | 0.0 | 0.0 | 0.8999 | 0.0 | 0.5436 | 0.3798 | 0.0 | nan | 0.0 | 0.2878 | 0.0 | 0.0 | 0.9485 | 0.8388 | 0.9598 | 0.0 | 0.0 | 0.3145 | 0.0 | nan | 0.7336 | 0.8788 | 0.6094 | 0.7062 | 0.3966 | nan | 0.3854 | 0.4897 | 0.0 | 0.7823 | 0.0 | nan | nan | 0.0 | 0.2782 | 0.0 | 0.0 | 0.7148 | 0.0 | 0.4182 | 0.3304 | 0.0 | nan | 0.0 | 0.2324 | 0.0 | 0.0 | 0.8415 | 0.7356 | 0.9130 | 0.0 | 0.0 | 0.2491 | 0.0 | | 0.6115 | 16.82 | 1800 | 0.5528 | 0.3266 | 0.3849 | 0.8539 | nan | 0.8521 | 0.9611 | 0.6840 | 0.8291 | 0.5057 | nan | 0.5070 | 0.7165 | 0.0 | 0.9267 | 0.0 | nan | nan | 0.0 | 0.3659 | 0.0 | 0.0 | 0.9007 | 0.0 | 0.5844 | 0.3961 | 0.0 | nan | 0.0 | 0.2827 | 0.0 | 0.0 | 0.9517 | 0.8371 | 0.9602 | 0.0 | 0.0 | 0.2848 | 0.0 | nan | 0.7414 | 0.8721 | 0.6312 | 0.7245 | 0.3979 | nan | 0.3987 | 0.4932 | 0.0 | 0.7799 | 0.0 | nan | nan | 0.0 | 0.2788 | 0.0 | 0.0 | 0.7242 | 0.0 | 0.4542 | 0.3464 | 0.0 | nan | 0.0 | 0.2326 | 0.0 | 0.0 | 0.8384 | 0.7318 | 0.9141 | 0.0 | 0.0 | 0.2386 | 0.0 | | 0.4766 | 18.69 | 2000 | 0.5449 | 0.3292 | 0.3907 | 0.8555 | nan | 0.8585 | 0.9611 | 0.7673 | 0.8223 | 0.5127 | nan | 0.4937 | 0.7164 | 0.0 | 0.9332 | 0.0 | nan | nan | 0.0 | 0.3858 | 0.0 | 0.0 | 0.9040 | 0.0 | 0.5848 | 0.4417 | 0.0 | nan | 0.0 | 0.3156 | 0.0 | 0.0 | 0.9413 | 0.8456 | 0.9600 | 0.0 | 0.0 | 0.2780 | 0.0 | nan | 0.7447 | 0.8755 | 0.6244 | 0.7325 | 0.3997 | nan | 0.3974 | 0.4985 | 0.0 | 0.7798 | 0.0 | nan | nan | 0.0 | 0.2904 | 0.0 | 0.0 | 0.7233 | 0.0 | 0.4555 | 0.3734 | 0.0 | nan | 0.0 | 0.2484 | 0.0 | 0.0 | 0.8451 | 0.7346 | 0.9161 | 0.0 | 0.0 | 0.2359 | 0.0 |