openvinotoolkit / open_model_zoo

Pre-trained Deep Learning models and demos (high quality and extremely fast)
https://docs.openvino.ai/latest/model_zoo.html
Apache License 2.0
4.1k stars 1.37k forks source link

Explanation needed for ANPR output #898

Closed aswin-datakalp closed 4 years ago

aswin-datakalp commented 4 years ago

I am passing an image which gives the output

array([[[[66.]],

    [[53.]],

    [[ 3.]],

    [[ 3.]],

    [[45.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]],

    [[-1.]]]], dtype=float32)

The problem is how to debug this output and come to a conclusion .

The explanation provide in the below home page is not adding up :

https://docs.openvinotoolkit.org/latest/_models_intel_license_plate_recognition_barrier_0001_description_license_plate_recognition_barrier_0001.html

Thanks

eaidova commented 4 years ago

In openvinio models output has fixed shape, while the decoded sequence can be less than full length. -1 used for empty places. The first -1 in the sequence used as stop-symbol (model does not support spaces) So the following output can be interpreted as WJ33B using symbols described in vocab dictionary from docs: https://docs.openvinotoolkit.org/latest/_models_intel_license_plate_recognition_barrier_0001_description_license_plate_recognition_barrier_0001.html

IRDonch commented 4 years ago

Closing, since the question was answered.