Why is Google Cloud Vision API Landmark Detection Returning Full Image Bounding Boxes?

I'm developing an application using the Google Cloud Vision API for landmark detection, but I'm encountering an issue where the returned bounding boxes always cover the entire image. For instance, for a 100x100 image, the bounding box is reported as [(0,0), (100, 0), (100, 100), (0, 100)], even when multiple, well-separated landmarks appear in the image.

I tried using the example image gs://cloud-samples-data/vision/landmark/st_basils.jpeg, and according to the documentation, I expected an output like:

JSONCopied
{
  "responses": [
    {
      "landmarkAnnotations": [
        {
          "mid": "/m/014lft",
          "description": "Saint Basil's Cathedral",
          "score": 0.7840959,
          "boundingPoly": {
            "vertices": [
              {
                "x": 812,
                "y": 1058
              },
              {
                "x": 2389,
                "y": 1058
              },
              {
                "x": 2389,
                "y": 3052
              },
              {
                "x": 812,
                "y": 3052
              }
            ]
          },
          "locations": [
            {
              "latLng": {
                "latitude": 55.752912,
                "longitude": 37.622315883636475
              }
            }
          ]
        }
      ]
    }
  ]
}

However, when I actually submit the request, I get the following output, where the bounding polygon again covers the entire image:

JSONCopied
{
  "responses": [
    {
      "landmarkAnnotations": [
        {
          "mid": "/m/0hm_7",
          "description": "Red Square",
          "score": 0.7341708,
          "boundingPoly": {
            "vertices": [
              {},
              {
                "x": 2487
              },
              {
                "x": 2487,
                "y": 3213
              },
              {
                "y": 3213
              }
            ]
          },
          "locations": [
            {
              "latLng": {
                "latitude": 55.753930299999993,
                "longitude": 37.620794999999994
              }
            }
          ]
        }
      ]
    }
  ]
}

I also watched this tutorial (https://www.youtube.com/watch?v=FZsdFvJLoa0), where the live results were promising, but my tests with the same image yielded no effective detection of the landmark.

My questions are:

• Has there been any significant change to the Cloud Vision API that affects landmark detection?
• Are there any settings or approaches to improve the bounding boxes?

Any insights or suggestions would be appreciated!