• [BUG SDK] neural network input wrong

Dear community!

It appears i get this error when i run an example of yours. It used to work tho, and i upgraded the depthai-sdk from 1.9.4 to 1.9.5. However i get this error now, smth is wrong with the neural network input data as it seems.

reference to your example : https://github.com/luxonis/depthai-experiments/blob/master/gen2-pedestrian-reidentification/main.py

my code produced then the same error obviously

[18443010B18E060F00] [20.4] [45.469] [NeuralNetwork(3)] [error] Input tensor 'data' (0) exceeds available data range. Data size (49152B), tensor offset (0), size (98304B) - skipping inference

    Hey gmaranta

    I checked and it works on my end. Perhaps you modified the models used. You will typically get the "input tensor data 0" error when compiling a blob for FP16 when it should be with U8.

    Hope this helps,
    Jaka

      The output from the network seemed to be different indeed. There is no functionality of the nnData.getFristLayerFP16(), somehow. But i did not change anything on the model itself. I could therefore remove the blob and install it again. But from the config settings etc it seems to be the same. Could it be a hardware problem, since it is running on yours?

        Hi gmaranta
        Did you pass the -ip U8 parameter when converting the blob?

        Thanks,
        Jaka