3dcnn-vis

Visualizing activations of 3D convolutional filters using keras-vis library.

Model architecture

Layer (type)	Output Shape	Param
conv1 (Conv3D)	(None, 16, 112, 112, 64)	5248
pool1 (MaxPooling3D)	(None, 16, 56, 56, 64)	0
conv2 (Conv3D)	(None, 16, 56, 56, 128)	221312
pool2 (MaxPooling3D)	(None, 8, 28, 28, 128)	0
conv3a (Conv3D)	(None, 8, 28, 28, 256)	884992
conv3b (Conv3D)	(None, 8, 28, 28, 256)	1769728
pool3 (MaxPooling3D)	(None, 4, 14, 14, 256)	0
conv4a (Conv3D)	(None, 4, 14, 14, 512)	3539456
conv4b (Conv3D)	(None, 4, 14, 14, 512)	7078400
pool4 (MaxPooling3D)	(None, 2, 7, 7, 512)	0
conv5a (Conv3D)	(None, 2, 7, 7, 512)	7078400
conv5b (Conv3D)	(None, 2, 7, 7, 512)	7078400
zero_padding3d_2 (ZeroPadding)	(None, 2, 9, 9, 512)	0
pool5 (MaxPooling3D)	(None, 1, 4, 4, 512)	0
flatten_2 (Flatten)	(None, 8192)	0
fc6 (Dense)	(None, 4096)	33558528
dropout_3 (Dropout)	(None, 4096)	0
fc7 (Dense)	(None, 4096)	16781312
dropout_4 (Dropout)	(None, 4096)	0
fc8 (Dense)	(None, 487)	1995239

Weights

Possible to use the pre-trained model in Caffe format or convert it to Keras format or simply download model weights in Keras format from here.

3D CNN cativations of filters

conv1

conv2 conv3a conv3b conv4a conv4b conv5a conv5b