Error Caused by: Layer 'cnn|concatenate': Input size mismatch. Size of input to this layer is different from the expected

4 次查看（过去 30 天）

显示更早的评论

Sarang Jokhio 2022-4-11

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1693915-error-caused-by-layer-cnn-concatenate-input-size-mismatch-size-of-input-to-this-layer-is-diffe

评论： Dmitry Rogachev 2023-3-20

在 MATLAB Online 中打开

Hi everyone,

I am trying to train a pointpiller network using code and functions from this example:

https://ww2.mathworks.cn/help/deeplearning/ug/lidar-object-detection-using-pointpillars-deep-learning.html

However, at the final stage,

[detector,info] = trainPointPillarsObjectDetector(cdsAugmented,detector,options);

I am receiving following error,

Processing data in minibatchqueue....
*************************************************************************
Data processing complete.
Error using dlnetwork/initialize
Invalid network.
Error in trainPointPillarsObjectDetector (line 182)
    dlnet = initialize(dlnet);
Error in TrainingPointPiller (line 231)
    [detector,info] = trainPointPillarsObjectDetector(cdsAugmented,detector,options);
Caused by:
    Layer 'cnn|concatenate': Input size mismatch. Size of input to this layer is different from the expected
    input size.
    Inputs to this layer:
        from layer 'cnn|up1|bn' (size 289(S) × 248(S) × 128(C) × 1(B))
        from layer 'cnn|up2|bn' (size 290(S) × 248(S) × 128(C) × 1(B))
        from layer 'cnn|up3|bn' (size 292(S) × 248(S) × 128(C) × 1(B))

Could you please point me to where I am making a mistake?

P.S. The above example was to train a point piller on two class (car and Truck), however, I modiefied the code and training it on just one class i.e. Car.

Thank you very much for your time :)