我们假设我们有两个边界框坐标张量:ViewData["INT_CertificationCategoriesID"] = new SelectList(_context.INT_CertificationCategories, "ID", "Category");
ViewData["INT_CertificationConferredID"] = new SelectList(_context.INT_CertificationConferred, "ID", "ConferredBy");
ViewData["INT_CertificationsID"] = new SelectList(_context.INT_Certifications, "ID", "Certification").Where(i => i.CategoryID.Contains(_context.INT_CertificationCategories.ID));
ViewData["RIM_ResourceID"] = new SelectList(_context.RIM_Resource, "ID", "FirstName");
,尺寸为[n,4],N
,尺寸为[k,4]。每个张量的每一行代表一个边界框的x1,y1,x2和y2。
Tensorflow中是否有一种有效的方法可以生成[{1}}的[n,k]矩阵K
?理想情况下,M
将是联盟的交叉点,但如果使用另一种方法也可以更简单。
答案 0 :(得分:0)
官方对象检测模型的source code具有计算iou的功能,请参阅函数Private Sub Worksheet_Change(ByVal Target As Excel.Range)
Select Case Target.Column
Case 1, 2, 3, 4
Select Case Target.Row Mod 6
Case 3, 4
Cells(6 * ((Target.Row - 3) \ 6) + 3, "F").Resize(5, 4).Select
End Select
Case 6, 7, 8, 9
Select Case Target.Row Mod 6
Case 2
Case Else
Cells(6 * ((Target.Row - 3) \ 6) + 3, "K").Resize(5, 4).Select
End Select
Case 11, 12, 13, 14
Select Case Target.Row Mod 6
Case 2
Case Else
Cells(6 * ((Target.Row - 3) \ 6) + 9, "A").Resize(2, 4).Select
End Select
End Select
End Sub
和intersection
以供参考