Debugging the SDK

The SDK looks like a black box and it may look that it’s hard to understand why may be the issue if it fails to recognize an image.

Here are some good practices to help you:
  1. Set the debug level to “verbose” and filter the logs with the keyword “doubango”: "debug_level": "verbose"

  2. Maybe the input image has the wrong size or format or we’re messing with it. To check how the input image looks like just before being forward to the neural networks enable dumping and set a path to the dump folder: "debug_write_input_image_enabled": true, "debug_internal_data_path": "<path to dump folder>".
    Check the sample applications to see how to generate a valid dump folder. The image will be saved on the device as “ultimateALPR-input.png” and to pull it from the device to your desktop use adb tool like this: adb pull <path to dump folder>/ultimateALPR-input.png

  3. The computer vision part is open source and you can match the lines on the logs to https://github.com/DoubangoTelecom/compv

The configuration should look like this:

{
    "debug_level": "verbose",
    "debug_write_input_image_enabled": true,
    "debug_internal_data_path": "<path to dump folder>"
}