HarisIqbal88 / PlotNeuralNet

Latex code for making neural networks diagrams
MIT License
21.54k stars 2.83k forks source link

I need help with my Architecture, plz #154

Open Bennetash opened 2 weeks ago

Bennetash commented 2 weeks ago

I am trying to make 2 parallel convolutional architectures with a fully connected fusion layer and an attention block for data. I have this code, could someone help me please?

mport sys sys.path.append('../') from pycore.tikzeng import *

Definir tu arquitectura

arch = [ to_head('..'), to_cor(), to_begin(),

# Entrada de la imagen
to_input('input_image.jpg', width=5, height=5, name='input'),

# Primera rama convolucional
to_Conv("conv1a", 32, 32, offset="(0,0,0)", to="(input-east)", height=32, depth=32, width=2),
to_Pool("pool1a", offset="(0,0,0)", to="(conv1a-east)", height=16, depth=16, width=1),

# Segunda rama convolucional
to_Conv("conv1b", 32, 32, offset="(1,0,0)", to="(input-east)", height=32, depth=32, width=2),
to_Pool("pool1b", offset="(0,0,0)", to="(conv1b-east)", height=16, depth=16, width=1),

# Conexión a la capa completamente conectada
to_Fc("fc", 64, offset="(2,0,0)", to="(pool1a-east)", height=1, depth=1, width=4, caption="Fully Connected"),
to_connection("pool1a", "fc"),
to_connection("pool1b", "fc"),

# Bloque de atención para datos de sensores
to_input('sensor_data.jpg', width=5, height=5, to="(input-south)", name='sensor_input'),
to_Attention("attention", 16, offset="(0,-3,0)", to="(sensor_input-east)", height=16, depth=16, width=4, caption="Attention Block"),
to_connection("sensor_input", "attention"),

# Capa de salida
to_SoftMax("soft1", 10, "(4,0,0)", "(fc-east)", caption="Output"),
to_connection("fc", "soft1"),

to_end()

]

def main(): file_name = str(sys.argv[0]).split('.')[0] to_generate(arch, file_name + '.tex')

if name == 'main': main()

I am basing it on the script example that is in the repository