Problem Raised when I use the size of width512, and height 256

[alanyannick]

Hi thx for your nice demo repo, just wondering whether you are familiar with the problem when I set the input width as 512 and height as 256(training size) for the ImageSegmentation demo (The demo works on 512x512.).

Out: Thread 1: EXC_BAD_ACCESS (code=1, address=0x0)

ImageSegmentation[78619:5076149] torch.cat(): Sizes of tensors must match except in dimension 1. Got 16 and 12 in dimension 3 (The offending index is 3)
  
  debug_handle:-1
  
Exception raised from check_cat_shape_except_dim at /Users/distiller/project/aten/src/ATen/native/TensorShape.cpp:114 (most recent call first):
frame #0: _ZN3c105ErrorC1ENS_14SourceLocationENSt3__112basic_stringIcNS2_11char_traitsIcEENS2_9allocatorIcEEEE + 75 (0x10a5479f1 in ImageSegmentation)
frame #1: _ZN3c106detail14torchCheckFailEPKcS2_jRKNSt3__112basic_stringIcNS3_11char_traitsIcEENS3_9allocatorIcEEEE + 106 (0x10a5465cd in ImageSegmentation)
frame #2: _ZN2at6native12_cat_out_cpuEN3c108ArrayRefINS_6TensorEEExRS3_ + 3129 (0x10a28e803 in ImageSegmentation)
frame #3: _ZN2at6native8_cat_cpuEN3c108ArrayRefINS_6TensorEEEx + 162 (0x10a28ec54 in ImageSegmentation)
frame #4: _ZN2at12_GLOBAL__N_112_GLOBAL__N_112wrapper__catEN3c108ArrayRefINS_6TensorEEEx + 14 (0x1096cfe35 in ImageSegmentation)
frame #5: _ZN3c104impl28wrap_kernel_functor_unboxed_INS0_6detail31WrapFunctionIntoRuntimeFunctor_IPFN2at6TensorENS_8ArrayRefIS5_EExES5_NS_4guts8typelist8typelistIJS7_xEEEEES8_E4callEPNS_14OperatorKernelENS_14DispatchKeySetES7_x + 24 (0x10984f82c in ImageSegmentation)
frame #6: _ZNK3c1010Dispatcher4callIN2at6TensorEJNS_8ArrayRefIS3_EExEEET_RKNS_19TypedOperatorHandleIFS6_DpT0_EEES9_ + 148 (0x109494b5a in ImageSegmentation)
frame #7: _ZN2at4_catEN3c108ArrayRefINS_6TensorEEEx + 65 (0x109449cfe in ImageSegmentation)
frame #8: _ZN2at6native3catEN3c108ArrayRefINS_6TensorEEEx + 3155 (0x10a28fb8e in ImageSegmentation)
frame #9: _ZN2at12_GLOBAL__N_112_GLOBAL__N_111wrapper_catEN3c108ArrayRefINS_6TensorEEEx + 14 (0x10977be54 in ImageSegmentation)
frame #10: _ZN3c104impl34call_functor_with_args_from_stack_INS0_6detail31WrapFunctionIntoRuntimeFunctor_IPFN2at6TensorENS_8ArrayRefIS5_EExES5_NS_4guts8typelist8typelistIJS7_xEEEEELb0EJLm0ELm1EEJS7_xEEENSt3__15decayINSA_21infer_function_traitsIT_E4type11return_typeEE4typeEPNS_14OperatorKernelENS_14DispatchKeySetEPNSF_6vectorINS_6IValueENSF_9allocatorISR_EEEENSF_16integer_sequenceImJXspT1_EEEEPNSC_IJDpT2_EEE + 182 (0x1097a88fe in ImageSegmentation)
frame #11: _ZN3c104impl31make_boxed_from_unboxed_functorINS0_6detail31WrapFunctionIntoRuntimeFunctor_IPFN2at6TensorENS_8ArrayRefIS5_EExES5_NS_4guts8typelist8typelistIJS7_xEEEEELb0EE4callEPNS_14OperatorKernelERKNS_14OperatorHandleENS_14DispatchKeySetEPNSt3__16vectorINS_6IValueENSM_9allocatorISO_EEEE + 24 (0x1097284d0 in ImageSegmentation)
frame #12: _ZNK3c1010Dispatcher9callBoxedERKNS_14OperatorHandleEPNSt3__16vectorINS_6IValueENS4_9allocatorIS6_EEEE + 119 (0x10a438cc1 in ImageSegmentation)
frame #13: _ZN5torch3jit6mobile16InterpreterState3runERNSt3__16vectorIN3c106IValueENS3_9allocatorIS6_EEEE + 5049 (0x10a44592f in ImageSegmentation)
frame #14: _ZNK5torch3jit6mobile8Function3runERNSt3__16vectorIN3c106IValueENS3_9allocatorIS6_EEEE + 160 (0x10a4370dc in ImageSegmentation)
frame #15: _ZNK5torch3jit6mobile6Method3runERNSt3__16vectorIN3c106IValueENS3_9allocatorIS6_EEEE + 766 (0x10a4490b8 in ImageSegmentation)
frame #16: _ZNK5torch3jit6mobile6MethodclENSt3__16vectorIN3c106IValueENS3_9allocatorIS6_EEEE + 24 (0x10a4497aa in ImageSegmentation)
frame #17: _ZN5torch3jit6mobile6Module7forwardENSt3__16vectorIN3c106IValueENS3_9allocatorIS6_EEEE + 176 (0x10a4e9b70 in ImageSegmentation)
frame #18: -[TorchModule + + (0x10a4e8f9e in ImageSegmentation)
frame #19: $s17ImageSegmentation14ViewControllerC7doInferyyypFyycfU_ + 1505 (0x10a4f44e1 in ImageSegmentation)
frame #20: $sIeg_IeyB_TR + 48 (0x10a4f4fc0 in ImageSegmentation)
frame #21: _dispatch_call_block_and_release + 12 (0x10bc168ac in libdispatch.dylib)
frame #22: _dispatch_client_callout + 8 (0x10bc17a88 in libdispatch.dylib)
frame #23: _dispatch_queue_override_invoke + 1032 (0x10bc19f06 in libdispatch.dylib)
frame #24: _dispatch_root_queue_drain + 351 (0x10bc295b6 in libdispatch.dylib)
frame #25: _dispatch_worker_thread2 + 135 (0x10bc29f1b in libdispatch.dylib)
frame #26: _pthread_wqthread + 220 (0x7fff5dcd89f7 in libsystem_pthread.dylib)
frame #27: start_wqthread + 15 (0x7fff5dcd7b77 in libsystem_pthread.dylib)

I guess some problems might raise during going through the elements of output(self.imageHelper.convertRGBBuffer). Any ideas to solve it? Thanks!