人員 偵測 (4.0 版)

影像分析 4.0 版提供偵測影像中出現人員的能力。 系統會傳回每個偵測到人員的周框方塊座標,以及信賴分數。

重要

我們藉由增強人員偵測案例的物件偵測模型來建置此模型。 人員 偵測並不涉及區分一張臉部與另一張臉部、預測或分類臉部屬性,或建立臉部範本(從代表臉部獨特特徵的影像產生的一組唯一數位)。

人員 偵測範例

下列 JSON 回應說明在根據視覺功能描述範例影像時,「分析 4.0 API」傳回的內容。

Photo of four people.

{
  "modelVersion": "2024-02-01",
  "metadata": {
    "width": 300,
    "height": 231
  },
  "peopleResult": {
    "values": [
      {
        "boundingBox": {
          "x": 0,
          "y": 41,
          "w": 95,
          "h": 189
        },
        "confidence": 0.9474349617958069
      },
      {
        "boundingBox": {
          "x": 204,
          "y": 96,
          "w": 95,
          "h": 134
        },
        "confidence": 0.9470965266227722
      },
      {
        "boundingBox": {
          "x": 53,
          "y": 20,
          "w": 136,
          "h": 210
        },
        "confidence": 0.8943784832954407
      },
      {
        "boundingBox": {
          "x": 170,
          "y": 31,
          "w": 91,
          "h": 199
        },
        "confidence": 0.2713555097579956
      }
    ]
  }
}

使用 API

人員偵測功能是分析影像 4.0 API一部分。 包含在People功能查詢參數中。 然後,當您取得完整的 JSON 回應時,剖析區段內容的 "people" 字串。

下一步