convolution - Fully-connected layer weight dimensions in TensorFlow ConvNet -

-

i've been coding along this example of convolution net in tensorflow , i'm mystified allocation of weights:

weights = {  # 5x5 conv, 1 input, 32 outputs 'wc1': tf.variable(tf.random_normal([5, 5, 1, 32])),  # 5x5 conv, 32 inputs, 64 outputs 'wc2': tf.variable(tf.random_normal([5, 5, 32, 64])),   # connected, 7*7*64 inputs, 1024 outputs 'wd1': tf.variable(tf.random_normal([7*7*64, 1024])),   # 1024 inputs, 10 outputs (class prediction) 'out': tf.variable(tf.random_normal([1024, n_classes]))   }

how know 'wd1' weight matrix should have 7 x 7 x 64 rows?

it's later used reshape output of second convolution layer:

# connected layer # reshape conv2 output fit dense layer input dense1 = tf.reshape(conv2, [-1, _weights['wd1'].get_shape().as_list()[0]])   # relu activation dense1 = tf.nn.relu(tf.add(tf.matmul(dense1, _weights['wd1']), _biases['bd1']))

by math, pooling layer 2 (conv2 output) has 4 x 4 x 64 neurons.

why reshaping [-1, 7*7*64]?

working start:

the input, _x of size [28x28x1] (ignoring batch dimension). 28x28 greyscale image.

the first convolutional layer uses padding=same, outputs 28x28 layer, passed max_pool k=2, reduces each dimension factor of two, resulting in 14x14 spatial layout. conv1 has 32 outputs -- full per-example tensor [14x14x32].

this repeated in conv2, has 64 outputs, resulting in [7x7x64].

tl;dr: image starts 28x28, , each maxpool reduces factor of 2 in each dimension. 28/2/2 = 7.


Comments

Post a Comment

Popular posts from this blog

ruby - Trying to change last to "x"s to 23 -

-
i have been playing around irb , have string looks this: irb(main):072:0> puts ["[\"4354 5432 5432 xxxxx\", \"6547 6547 8543 xxxxx\", \"2344 6543 6674 xxxxx\", \"2346 6236 7543 xxxxx\", \"1273 5585 5587 xxxxx\"]"] => nil irb(main):073:0> what want gsub last 2 "x" s of each 17 digit combination, 23 . what i've tried far (i've been using rubular): a.gsub(/\d[x]/,"23") <= grabs "x" s i'm close a.gsub(/\w[x{2}]/,"23") <= grabs "x" s grabs two digits each combination. is there easier way i'm not understanding? a.gsub(/\d{4} \d{4} \d{4} xxx\kxx/, '23') \d{4} = 4 digits what \k drop matched far. can use don't replace entire number, still able validate have in desired format.
Read more

jquery - Clone last and append item to closest class -

-
this bare-bones code on page: <div class="list-container"> <div class="single-list-item"> <button>remove</button> </div> <div class="single-list-item"> <button>remove</button> </div> <div class="single-list-item"> <button>remove</button> </div> </div> <p><button class="add-list-item">add</button></p> <div class="list-container"> <div class="single-list-item"> <button>remove</button> </div> <div class="single-list-item"> <button>remove</button> </div> <div class="single-list-item"> <button>remove</button> </div> </div> <p><button class="add-list-item">add</button></p> <di...
Read more

css - Can I use the :after pseudo-element on an input field? -

-
i trying use :after css pseudo-element on input field, not work. if use span , works ok. <style type="text/css"> .mystyle:after {content:url(smiley.gif);} .mystyle {color:red;} </style> this works (puts smiley after "buu!" , before "some more") <span class="mystyle">buuu!</span>a more this not work - colors somevalue in red, there no smiley. <input class="mystyle" type="text" value="somevalue"> what doing wrong? should use pseudo-selector? note: cannot add span around input , because being generated third-party control. :after , :before not supported in internet explorer 7 , under, on elements. it's not meant used on replaced elements such form elements (inputs) , image elements . in other words it's impossible pure css. however if using jquery can use $(".mystyle").after("add smiley here"); api docs on .after to ...
Read more