You have 2 classes but you give 10 output from the logsoftmax.

from cvpr2015.

I changed the line net:add(nn.Linear(84, 10) to net:add(nn.Linear(84, 2), but the error remains.
Help. I'm new to this.

from cvpr2015.

Your batch size is very big. It could be the problem. Try with smaller batch.

from cvpr2015.

Does not work. could you help me with the correct data?
My goal is that the neural network can distinguish human from animal.

from cvpr2015.

net:add(nn.View(1699))

must be

net:add(nn.View(5184)).
for 181816.

I dont know what the problem is directly. It says something about batch size. Is this the full error message.

from cvpr2015.

New error!

nn.Sequential {
[input -> (1) -> (2) -> (3) -> (4) -> (5) -> (6) -> (7) -> (8) -> (9) -> (10) -> (11) -> (12) -> (13) -> output]
(1): nn.SpatialConvolution(3 -> 6, 18x18)
(2): nn.ReLU
(3): nn.SpatialMaxPooling(2x2, 2,2)
(4): nn.SpatialConvolution(6 -> 16, 18x18)
(5): nn.ReLU
(6): nn.SpatialMaxPooling(2x2, 2,2)
(7): nn.View(5184)
(8): nn.Linear(5184 -> 120)
(9): nn.ReLU
(10): nn.Linear(120 -> 84)
(11): nn.ReLU
(12): nn.Linear(84 -> 2)
(13): nn.LogSoftMax
}
[0.0371s]
th> criterion = nn.ClassNLLCriterion()
[0.0761s]
th>
[0.0000s]
th> trainer = nn.StochasticGradient(net, criterion)
[0.0001s]
th> trainer.learningRate = 0.001
[0.0000s]
th> trainer.maxIteration = 5
[0.0000s]
th> trainer:train(trainset)

StochasticGradient: training

.../facedetect/torch/install/share/lua/5.1/nn/Container.lua:67:
In 7 module of nn.Sequential:
/root/facedetect/torch/install/share/lua/5.1/nn/View.lua:47: input view (16x11x11) and desired view (5184) do not match
stack traceback:
[C]: in function 'error'
/root/facedetect/torch/install/share/lua/5.1/nn/View.lua:47: in function 'batchsize'
/root/facedetect/torch/install/share/lua/5.1/nn/View.lua:79: in function </root/facedetect/torch/install/share/lua/5.1/nn/View.lua:77>
[C]: in function 'xpcall'
.../facedetect/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
...facedetect/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
...ct/torch/install/share/lua/5.1/nn/StochasticGradient.lua:35: in function 'train'
[string "_RESULT={trainer:train(trainset)}"]:1: in main chunk
[C]: in function 'xpcall'
/root/facedetect/torch/install/share/lua/5.1/trepl/init.lua:661: in function 'repl'
...tect/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:199: in main chunk
[C]: at 0x004064f0

WARNING: If you see a stack trace below, it doesn't point to the place where this error occurred. Please use only the one above.
stack traceback:
[C]: in function 'error'
.../facedetect/torch/install/share/lua/5.1/nn/Container.lua:67: in function 'rethrowErrors'
...facedetect/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
...ct/torch/install/share/lua/5.1/nn/StochasticGradient.lua:35: in function 'train'
[string "_RESULT={trainer:train(trainset)}"]:1: in main chunk
[C]: in function 'xpcall'
/root/facedetect/torch/install/share/lua/5.1/trepl/init.lua:661: in function 'repl'
...tect/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:199: in main chunk
[C]: at 0x004064f0
[0.3159s]

it looked at the beginning:
th> net:add(nn.View(16x9x9)) not "net:add(nn.View(1699))"
th> net:add(nn.Linear(16x9x9, 120)) not "net:add(nn.Linear(1699, 120))"

from cvpr2015.

Ok. 16 x 11 x 11 would be correct.

from cvpr2015.

I replaced it and did from the beginning:

th> net:add(nn.SpatialConvolution(3, 6, 9, 9))
th> net:add(nn.SpatialConvolution(6, 16, 9, 9))
th> net:add(nn.View(16x9x9))
th> net:add(nn.Linear(16x9x9, 120))

net = nn.Sequential()
net:add(nn.SpatialConvolution(3, 6, 11, 11))
net:add(nn.ReLU())
net:add(nn.SpatialMaxPooling(2,2,2,2))
net:add(nn.SpatialConvolution(6, 16, 11, 11))
net:add(nn.ReLU())
net:add(nn.SpatialMaxPooling(2,2,2,2))
net:add(nn.View(16x11x11))
net:add(nn.Linear(16x11x11, 120))
net:add(nn.ReLU())
net:add(nn.Linear(120, 84))
net:add(nn.ReLU())
net:add(nn.Linear(84, 2))
net:add(nn.LogSoftMax())

Error:
th> trainer:train(trainset)

StochasticGradient: training

.../facedetect/torch/install/share/lua/5.1/nn/Container.lua:67:
In 7 module of nn.Sequential:
/root/facedetect/torch/install/share/lua/5.1/nn/View.lua:47: input view (16x16x16) and desired view (1936) do not match
stack traceback:
[C]: in function 'error'
/root/facedetect/torch/install/share/lua/5.1/nn/View.lua:47: in function 'batchsize'
/root/facedetect/torch/install/share/lua/5.1/nn/View.lua:79: in function </root/facedetect/torch/install/share/lua/5.1/nn/View.lua:77>
[C]: in function 'xpcall'
.../facedetect/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
...facedetect/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
...ct/torch/install/share/lua/5.1/nn/StochasticGradient.lua:35: in function 'train'
[string "_RESULT={trainer:train(trainset)}"]:1: in main chunk
[C]: in function 'xpcall'
/root/facedetect/torch/install/share/lua/5.1/trepl/init.lua:661: in function 'repl'
...tect/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:199: in main chunk
[C]: at 0x004064f0

WARNING: If you see a stack trace below, it doesn't point to the place where this error occurred. Please use only the one above.
stack traceback:
[C]: in function 'error'
.../facedetect/torch/install/share/lua/5.1/nn/Container.lua:67: in function 'rethrowErrors'
...facedetect/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
...ct/torch/install/share/lua/5.1/nn/StochasticGradient.lua:35: in function 'train'
[string "_RESULT={trainer:train(trainset)}"]:1: in main chunk
[C]: in function 'xpcall'
/root/facedetect/torch/install/share/lua/5.1/trepl/init.lua:661: in function 'repl'
...tect/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:199: in main chunk
[C]: at 0x004064f0

write 16x16x16
next error:
input view (16x12x12) and desired view (4096) do not match

maybe I'm not there to change the value?

from cvpr2015.

how to determine these parameters:

5x5 convolution kernel?
net:add(nn.SpatialConvolution(3, 6, 5, 5))

reshapes from a 3D tensor of 16x5x5 into 1D tensor of 1655?
net:add(nn.View(1655))

from cvpr2015.

Convolution makes your pictures smaller by m-1 n-1 where mxn is your filter. Subsampling makes a division over size. Your pictures are 96x96. 5x5 convolution will make them 92x92 if used without padding. Search about convolution and pooling.

from cvpr2015.