NCSDK V2.05 mvNCCheck failed when caffe network with 1 channel input

idata · ‎06-21-2018

the network has input like this：

name: "suma"
input: "data"
input_shape {
    dim: 1
    dim: 1
    dim: 11
    dim: 11
}

that means we just need 1 channel，not rgb 3 channels.

And the mvNCCheck failed:

Traceback (most recent call last):
  File "/usr/local/bin/ncsdk/mvNCCheck.py", line 202, in <module>
    quit_code = check_net(args.network, args.image, args.inputnode, args.outputnode, args.nshaves, args.inputsize, args.weights, args)
  File "/usr/local/bin/ncsdk/mvNCCheck.py", line 169, in check_net
    load_ret = load_network(args, parser, myriad_config)
  File "/usr/local/bin/ncsdk/Controllers/Scheduler.py", line 103, in load_network
    parse_ret = parse_caffe(arguments, myriad_conf)
  File "/usr/local/bin/ncsdk/Controllers/CaffeParser.py", line 414, in parse_caffe
    channel_swap=channel_swap)
  File "/usr/local/bin/ncsdk/Controllers/MiscIO.py", line 273, in preprocess_img
    data[0] = data[0][np.argsort(channel_swap), :, :]
IndexError: index 2 is out of bounds for axis 0 with size 1

I debug mvNCCheck.py, it caused by :

 input_data = preprocess_img(input_data,
            raw_scale=raw_scale,
            mean=mean,
            channel_swap=channel_swap)

when input_image is None or input_image == "Debug" in this situation, we needn't do channel_swap, bucause the input_image is gengerated by

input_data = np.random.uniform(-1, 1, input_shape).astype(dtype=data_type)

So I just modified channel_swap=channel_swap to channel_swap=None, the preprocess_img() can do mean and scale as well, which in NCSDK V2.04 it didn't do.

idata · ‎06-21-2018

@thomascatlee Thanks for reporting this. mvNCCheck is designed for 3 channel models at the moment so that could be the reason why you're running into issues with your 1 dim network. Your feedback is important and we will consider this for any future releases.

lius · ‎07-19-2019

hi, mdata

does the mvnc support the input channel of 1 now(gray image)?