2

I have six channel image (2 RGB image concatenated). I want to train the model with these images using AlexNet. I packed the images into lmdb. Then I used the OTHER option for dataset and model, however I am getting the following error when I am creating the model .

ERROR: Top blob 'data' produced by multiple sources. 
Creating layer train-data
Creating Layer train-data
Top blob 'data' produced by multiple sources.
Opened lmdb /lmdb_database/train_labels

I am using standard AlexNet architecture.

Data Preparation in lmdb

I have two images in RGB coming from two different modalities. I modified createdb.py script from siamese example to concatenate two images. Once I have the dataset then I am using the standard AlexNet Prototext to train the model

cpwah
  • 141
  • 1
  • 3
  • 13

2 Answers2

0

Check your input data layers, is it possible you forgot to specify stage/phase for them? it seems like caffe is trying to use both at the same time

Shai
  • 111,146
  • 38
  • 238
  • 371
  • I am using standard AlexNet prototxt, link mentioned above. I believe the AlexNet assume that the two concatenated images are coming from both train and val. Can someone explain to me how to use multiple input. @Shai – cpwah Apr 10 '17 at 11:24
  • 1
    @cpwah please describe in more detail how you prepare your data and how the input layer prototxt looks like – Shai Apr 10 '17 at 11:26
  • I have edited the question. Please let me know if you need any further information. – cpwah Apr 10 '17 at 11:30
0

Today I have encountered this problem too. And I found a solution. My data layer's name is 'data' and then top is 'data' too, so it has a wrong. And I change my data layer, it works.

user7860264
  • 1