convlstm.rar_ConvLSTM_ConvLstm、_conlstm实现分类_convlstm代码_卷积LSTM实现

# -*- coding: utf-8 -*- import tensorflow as tf import numpy as np import os from cell import ConvLSTMCell if 'session' in locals() and session is not None: print('Close interactive session') session.close() def main(): shape=[132, 228] kernel = [3, 3] filters = 12 batch_size=181 num_steps=8 channels=7 inputs = tf.placeholder(tf.float32, [batch_size, num_steps] + shape + [channels]) cell = ConvLSTMCell(shape, filters, kernel) outputs, state = tf.nn.dynamic_rnn(cell, inputs, dtype=inputs.dtype, time_major=True) print(inputs.shape) with tf.Session() as sess: sess.run(tf.global_variables_initializer()) for idx, epoch in enumerate(gen_epochs(num_epochs, num_steps)): for step, (X, Y) in enumerate(epoch): o, s = sess.run([outputs, state], feed_dict={inputs:inp}) print o.shape #init_state = tf.zeros([batch_size, state_size]) #path='/home/zhanglu/python/AIRnet/gfs_air'#gfs_air201506.npz' #gen_bath(path,batch_size,num_steps,channels,shape) ''' if os.path.exists(path): for root,dirs,files in os.walk(path): sub_f=[root+f for f in files if f[-3:] == 'npz'] for filename in sub_f: print(filename) with np.load(filename) as f_d: b_arr=np.array(f_d['gfs_air']) #0:tm,1:rh,2:ugrd,3:vgrd,4:prate,5:tcdc,6:CO,7:PM10,8:O3,9:SO2,10:NO2,11:AQI,12:PM25 #tmp:Temperature #rh :Relative Humidity #ugrd:U-Component of Wind #vgrd:V-Component of Wind #prate:Precipitation Rate #tcdc:Total Cloud Cover #data_atm=np.array(b_arr[:,:,:,0]) #data_PM25=np.array(b_arr[:,:,:,12]) #data=np.stack((np.array(b_arr[:,:,:,0]),np.array(b_arr[:,:,:,1]),np.array(b_arr[:,:,:,12]))) num+=b_arr.shape[2] print(num,b_arr.shape[2]) ''' def gen_bath(path,batch_size,num_steps,channels,shape): month_d=[239,248,240,248,248,240,248, 240,248,248,232,248,240,248, 240,248,248,240,248,240,248, 248,224,248,240,248,240,248,248,169] file_nameF=['201505','201506','201507','201508','201509','201510','201511','201512','201601', '201602','201603','201604','201605','201606','201607','201608','201609','201610', '201611','201612','201701','201702','201703','201704','201705','201706','201707', '201708','201709'] #H:\IEEE_ICDM\gfs_air201504.npz atm_type=12 #0:tm,1:rh,2:ugrd,3:vgrd,4:prate,5:tcdc,6:CO,7:PM10,8:O3,9:SO2,10:NO2,11:AQI,12:PM25 #path='H:\\IEEE_ICDM\\gfs_air' time_len=7240 batch_partition_length = time_len // batch_size print(batch_partition_length) data_x = np.zeros([batch_size, batch_partition_length] + shape + [channels], dtype=np.float32) data_y = np.zeros([batch_size, batch_partition_length] + shape + [channels], dtype=np.float32) print(data_x.shape) j=0 sum_time=month_d[0] x=month_d[0] filename=path+file_nameF[0]+'.npz' next_range=0 ii=0 for i in range(batch_size): x=x-batch_partition_length #print(i*batch_partition_length) if sum_time >i*batch_partition_length: b_ii=ii ii=ii+batch_partition_length if (sum_time-i*batch_partition_length)<= 40: range1=-sum_time+i*batch_partition_length ii=40+range1 if range1<0: if j+1 >=len(file_nameF): break filename1=path+file_nameF[j+1]+'.npz' print(filename,range1) print(filename1,range1+40) f=np.load(filename) b=np.array(f['gfs_air']) data_b=np.array(b[:,:,range1:,0:6]) data_b1=np.array(b[:,:,range1:,atm_type]) data_b1=data_b1[:,:,:,np.newaxis] data_b=np.concatenate((data_b,data_b1),axis=3) f.close() f=np.load(filename1) b=np.array(f['gfs_air']) data_a=np.array(b[:,:,0:range1+40,0:6]) data_a1=np.array(b[:,:,0:range1+40,atm_type]) data_a1=data_a1[:,:,:,np.newaxis] data_a=np.concatenate((data_a,data_a1),axis=3) f.close() data=np.concatenate((data_b,data_a),axis=2) #print('data1:',data.shape) #print('data_a:',data_a.shape) #print('data:',data.shape) else: a_ii=ii f=np.load(filename) b=np.array(f['gfs_air']) data=np.array(b[:,:,b_ii:a_ii,0:6]) data1=np.array(b[:,:,b_ii:a_ii,atm_type]) data1=data1[:,:,:,np.newaxis] data=np.concatenate((data,data1),axis=3) f.close() #print('data2:',data.shape) else: j=j+1 if j+1 > len(file_nameF): break sum_time=sum_time+month_d[j] filename=path+file_nameF[j]+'.npz' data=data.swapaxes(2,0) data=data.swapaxes(2,1) print('data:',data.shape) ''' #print(batch_partition_length) #with np.load(filename) as f_d # data_x[i] = raw_x[batch_partition_length * i:batch_partition_length * (i + 1)] # data_y[i] = raw_y[batch_partition_length * i:batch_partition_length * (i + 1)] #print(data_x.shape) global batch_size,timesteps,shape,kernel,channels,filters batch_size = 32 timesteps = 100 shape = [640, 480] kernel = [3, 3] channels = 6 filters = 12 # Create a placeholder for videos. inputs = tf.placeholder(tf.float32, [batch_size, timesteps] + shape + [channels],name='input_placeholder') outputs = tf.placeholder(tf.float32, [batch_size, timesteps] + shape + [1],name='labels_placeholder') print(tf.shape(inputs),inputs.shape) cell = ConvLSTMCell(shape, filters, kernel) outputs, state = tf.nn.dynamic_rnn(cell, inputs, dtype=inputs.dtype) ################################################ # Add the ConvLSTM step. cell = ConvLSTMCell(shape, filters, kernel) outputs, state = tf.nn.dynamic_rnn(cell, inputs, dtype=inputs.dtype) # There's also a ConvGRUCell that is more memory efficient. from cell import ConvGRUCell cell = ConvGRUCell(shape, filters, kernel) outputs, state = tf.nn.dynamic_rnn(cell, inputs, dtype=inputs.dtype) # It's also possible to enter 2D input or 4D input instead of 3D. shape = [100] kernel = [3] inputs = tf.placeholder(tf.float32, [batch_size, timesteps] + shape + [channels]) cell = ConvLSTMCell(shape, filters, kernel) outputs, state = tf.nn.bidirectional_dynamic_rnn(cell, cell, inputs, dtype=inputs.dtype) shape = [50, 50, 50] kernel = [1, 3, 5] inputs = tf.placeholder(tf.float32, [batch_size, timesteps] + shape + [channels]) cell = ConvGRUCell(shape, filters, kernel) outputs, state= tf.nn.bidirectional_dynamic_rnn(cell, cell, inputs, dtype=inputs.dtype) ''' if __name__ == '__main__': main() print('##FINISH##')