Image Search

Enhance the effectiveness of your image searches with these helpful guidelines for capturing photos. By following these recommendations, you can optimize the performance of image recognition capabilities within Physna. Discover examples of images that perform well and those to avoid, ensuring efficient and accurate asset discovery.

We continuously enhance our image recognition capabilities. Currently, to achieve optimal performance, consider the following properties:

  1. Images with the object by itself.
  2. Images with a plain background.
  3. The color of the object usually does not affect detection outcomes.

Image Guidance

Examples of Images that Perform Well

Images that perform well

  • plain background
  • object isolated on its own

Examples of Images to Avoid

Images to avoid

  • multiple objects
  • cluttered or busy background

Quick Start

Upload Images

Request Signed URLs

Much like uploading models, uploading images with which to search requires getting a signed URL first. This is accomplised with the POST /v2/images endpoint (spec).

The response will include the appropriate headers and upload URL that can be used to upload the image.

{
 "image": {
  "id": "<imageId>",
  "uploadUrl": "<upload-url>",
  "headers": {
    "<header1-name>": "<header1-value>",
    "<header2-name>": "<header2-value>",
  }
 }
}

The returned imageId will be used to perform the seach after the image is uploaded.

Note: The headers returned for the signed URL are dynamic, and that all headers are required in the request to send the image file to the signed URL.

Send Image Files

Once the signed URL is available, the image can be then uploaded via a PUT request including the file and the supplied headers.

Example using curl:

curl -X PUT "<upload-url>" \
 --upload-file local/path/to/image.png \
 -H '<header1-name>: <header1-value>' \
 -H '<header2-name>: <header2-value>' \

Perform Searches

Searches are performed with one or more images via the GET /v2/images/model-matches endpoint (spec). It takes a list of imageIds for already-uploaded images and will visually search for models that have completed ingestion.