The SOC dataset including 6K salient images (3K) and non salient images (3K).
SOC數據集包括6K突出圖像(3K)和非突出圖像(3K)。
=====================================================
We devided our dataset into TrainSet (training), ValSet (validation), and TestSet (test)
in the ratio of 6:2:2.
我們將數據集分爲TrainSet(訓練)、ValSet(驗證)和TestSet(測試)
比例是6:2:2。
Both TrainSet and ValSet consist of 'Imgs' folder, 'gt' folder and 'Instance gt' folder.
While TestSet only include the image for test only. You can evaluate you code
in our website: http://dpfan.net/ (you can also send email to me: [email protected])
TrainSet和ValSet都是由‘Imgs’文件夾、‘gt’文件夾和‘Instance gt’文件夾組成。
而TestSet只包含用於測試的圖像。您可以評估您的代碼
在我們的網站:http://dpfan.net/(你也可以發郵件給我:[email protected])
'Imgs' folder: contains original color images.
'gt' folder: contains object-level salient object ground truth.
'Instance gt' folder: including 3 subfolder ('Attribute' folder, 'gt' folder, 'Instance name' folder).
'Attribute folder': describe each challenging attribute in each images.
'gt' folder: contains the instance-level salient object ground truth.
'Instance name' folder: maintains the instance name of each labeled instance.
'Imgs'文件夾:包含原始彩色圖像。
“gt”文件夾:包含對象級別的突出對象地面真相。
“實例gt”文件夾:包括3個子文件夾(“屬性”文件夾、“gt”文件夾、“實例名”文件夾)。
“屬性文件夾”:在每張圖片中描述每個具有挑戰性的屬性。
“gt”文件夾:包含實例級突出對象地面真相。
“實例名”文件夾:維護每個標記實例的實例名。