[ BVLC_Alexnet Caffe Model, Torch] : Misclassification ??

[Pokkalla Harsha Vardhan]

Hello,
I have loaded bvlc_alexnet caffe model in Torch and classifying the images from ImageNet Dataset. I am unable to interpret the answers based on class labels. I am printing out the index of class label that has highest probability. I am getting '670' & '795' for most of the images. I have tried with more than 5 images per class (elephant, black olive, beer bottles, digital clock).
Would someone please help me interpreting these results. Also, would someone help me with decoding true labels of 1000 classes (bvlc_alexnet caffe model) corresponding to synset_words.txt (ImageNet)

Few of them are :

My code is as follows:

require 'clnn'
require 'ccn2'
require 'loadcaffe'
require 'nn'
require 'inn'
require 'cunn'
require 'image'
model = loadcaffe.load('deploy.prototxt.lua','bvlc_alexnet.caffemodel','ccn2')
model:evaluate()
--print(model)
--reading images of batch size 32
img = image.load('elephant.jpeg')
img_sca = image.scale(img,227,227)
imB =torch.CudaTensor(32,3,227,227)
for i = 1,32 do imB[i]=img_sca end

--passing it to model--
model:forward(imB)

--Classification--
op26 = model.modules[26].output[1]
max_op26, class_idx = torch.max(op26,1)
print(class_idx)

[Sergey Zagoruyko]

You are not preprocessing the images.

Check how it's done here https://github.com/torch/tutorials/blob/master/7_imagenet_classification/classify.lua

--
You received this message because you are subscribed to the Google Groups "torch7" group.
To unsubscribe from this group and stop receiving emails from it, send an email to torch7+un...@googlegroups.com.
To post to this group, send email to tor...@googlegroups.com.
Visit this group at http://groups.google.com/group/torch7.
For more options, visit https://groups.google.com/d/optout.

[Pokkalla Harsha Vardhan]

works now. Thanks a lot :)

On Wednesday, 28 October 2015 13:32:51 UTC-4, Sergey Zagoruyko wrote:

You are not preprocessing the images.

Check how it's done here https://github.com/torch/tutorials/blob/master/7_imagenet_classification/classify.lua