Running rec_test.m without pretraining

[ghost]

There are 4 files, discriminative_10_class.mat, discriminative_40_class.mat, generative_10_class.mat and generative_40_class.mat, which I think are pre-trained models. But when I run call rec_test.m with any of these models, I get the following error:

Error using feval Argument must contain a string or function_handle.

Error in myConvolve2 (line 32) target_gpu = feval(kConv,...

Error in rec_test (line 43) hidden_presigmoid = myConvolve2(kConv_forward2, batch_data, model.layers{l}.uw, stride, 'forward');

My question is whether pre-training using run_pretrain.m is necessary for running rec_test.m. Also, from where do we get the accuracy of the test data?

Thanks

[jawhster]

I also receive this error when trying to run sample_test_classification, even after running run_pretrain.m and passing in the trained model. Did you ever get an answer to this question? Also, how do I run the model to classify a given depth image (i.e. which file does this)? Thanks, Justin

[tr4B]

Have you received any answers?