colored displaying of weights

nn_ds_neibs1_tmp.py

@@ -250,10 +250,10 @@ files_train_hvar = ["/home/oleg/GIT/python3-imagej-tiff/data_sets/tf_data_rand2/
                     "/home/oleg/GIT/python3-imagej-tiff/data_sets/tf_data_rand2/train007_R1_GT_1.5.tfrecords",
 ]
 
-files_train_lvar = ["/home/oleg/GIT/python3-imagej-tiff/data_sets/tf_data_rand2/train000_R1_LE_1.5.tfrecords",
-                    ]
-files_train_hvar = ["/home/oleg/GIT/python3-imagej-tiff/data_sets/tf_data_rand2/train000_R1_GT_1.5.tfrecords",
-]
+#files_train_lvar = ["/home/oleg/GIT/python3-imagej-tiff/data_sets/tf_data_rand2/train000_R1_LE_1.5.tfrecords",
+#                    ]
+#files_train_hvar = ["/home/oleg/GIT/python3-imagej-tiff/data_sets/tf_data_rand2/train000_R1_GT_1.5.tfrecords",
+#]
 #files_train_hvar = []
 #file_test_lvar=     "/home/eyesis/x3d_data/data_sets/tf_data_3x3a/train000_R1_LE_1.5.tfrecords" # "/home/eyesis/x3d_data/data_sets/train-000_R1_LE_1.5.tfrecords"
 file_test_lvar=     "/home/oleg/GIT/python3-imagej-tiff/data_sets/tf_data_rand2/testTEST_R1_LE_1.5.tfrecords"
@@ -855,18 +855,18 @@ with tf.Session()  as sess:
 
     merged = tf.summary.merge_all()
     
-    vis_placeholder = tf.placeholder(tf.float32, [1,32,325,3])
-    some_image2 = tf.summary.image('custom_test', vis_placeholder)
+    # display weights, part 1 begin
+    import numpy_image_test as nit
     
     l1 = NN_LAYOUT1.index(next(filter(lambda x: x!=0, NN_LAYOUT1)))
     l2 = NN_LAYOUT2.index(next(filter(lambda x: x!=0, NN_LAYOUT2)))
-    with tf.variable_scope('g_fc_sub'+str(l1),reuse=tf.AUTO_REUSE):
-        w = tf.get_variable('weights',shape=[325,32])
-        wd = w[...,tf.newaxis]
-        wds = tf.stack([wd]*3,axis=0)
-        #print(wd.shape)
-        #some_image = tf.summary.image("tfsi_test",wds.eval(),max_outputs=1)
-        some_image = tf.summary.image("tfsi_test",wds,max_outputs=1)
+    
+    wimg1_placeholder = tf.placeholder(tf.float32, [1,160,80,3])
+    wimg1 = tf.summary.image('weights/sub_'+str(l1), wimg1_placeholder)
+    
+    wimg2_placeholder = tf.placeholder(tf.float32, [1,120,60,3])
+    wimg2 = tf.summary.image('weights/inter_'+str(l2), wimg2_placeholder)
+    # display weights, part 1 end
         
     train_writer =    tf.summary.FileWriter(TRAIN_PATH, sess.graph)
     test_writer  =    tf.summary.FileWriter(TEST_PATH, sess.graph)
@@ -970,23 +970,27 @@ with tf.Session()  as sess:
 #        _,_=sess.run([tf_ph_G_loss,tf_ph_sq_diff],feed_dict={tf_ph_G_loss:test_avg, tf_ph_sq_diff:test2_avg})
             
         #train_writer.add_summary(some_image.eval(), epoch)
+        
+        # display weights, part 2 begin
         l1 = NN_LAYOUT1.index(next(filter(lambda x: x!=0, NN_LAYOUT1)))
         l2 = NN_LAYOUT2.index(next(filter(lambda x: x!=0, NN_LAYOUT2)))
-        with tf.variable_scope('g_fc_sub'+str(l1),reuse=tf.AUTO_REUSE):
-            w = tf.get_variable('weights',shape=[325,32])
-            wd = w[tf.newaxis,...]
-            wds = tf.stack([wd]*3,axis=-1)
         
-            timg_min = tf.reduce_min(w).eval()
-            timg_max = tf.reduce_max(w).eval()
+        with tf.variable_scope('g_fc_sub'+str(l1),reuse=tf.AUTO_REUSE):
+            w = tf.get_variable('weights',shape=[325,NN_LAYOUT1[l1]])
+            w = tf.transpose(w,(1,0))            
+            img1 = nit.tiles(nit.coldmap(w.eval(),zero_span=0.0002),(1,4,9,9),tiles_per_line=2,borders=True)
+            img1 = img1[np.newaxis,...]
             
-            timg = wds.eval()
+        train_writer.add_summary(wimg1.eval(feed_dict={wimg1_placeholder: img1}), epoch)
 
-            timg[:,:,:,0] = timg_min
-            timg[:,:,:,1] = timg_min
-            timg = np.transpose(timg,(0,2,1,3))
+        with tf.variable_scope('g_fc_inter'+str(l2),reuse=tf.AUTO_REUSE):
+            w = tf.get_variable('weights',shape=[144,NN_LAYOUT1[l2]])
+            w = tf.transpose(w,(1,0))            
+            img2 = nit.tiles(nit.coldmap(w.eval(),zero_span=0.0002),(3,3,4,4),tiles_per_line=4,borders=True)
+            img2 = img2[np.newaxis,...]
             
-        train_writer.add_summary(some_image2.eval(feed_dict={vis_placeholder: timg}), epoch)
+        train_writer.add_summary(wimg2.eval(feed_dict={wimg2_placeholder: img2}), epoch)
+        # display weights, part 2 end
 
         train_writer.add_summary(train_summary, epoch)
         test_writer.add_summary(test_summaries[0], epoch)

numpy_image_test.py

+#!/usr/bin/env python3
+
+import numpy as np
+import matplotlib.pyplot as plt
+import math
+
+def hamming_window(x,N):
+  y = 0.2 - 0.46*math.cos(2*math.pi*x/(N-1))
+  return y
+
+# input: np.array(a,b) - 1 channel
+# output: np.array(a,b,3) - 3 color channels
+def coldmap(img,zero_span=0.2):
+
+  out = np.dstack(3*[img])
+
+  img_min = np.nanmin(img)
+  img_max = np.nanmax(img)
+
+  #print("min: "+str(img_min)+", max: "+str(img_max))
+
+  ch_r = out[...,0]
+  ch_g = out[...,1]
+  ch_b = out[...,2]
+
+  # blue for <0
+  ch_r[img<0] = 0
+  ch_g[img<0] = 0
+  ch_b[img<0] = -ch_b[img<0]
+
+  # red for >0
+  ch_r[img>0] = ch_b[img>0]
+  ch_g[img>0] = 0
+  ch_b[img>0] = 0
+
+  # green for 0
+  ch_r[img==0] = 0
+  ch_g[img==0] = img_max
+  ch_b[img==0] = 0
+
+  # green for zero vicinity
+  ch_r[abs(img)<zero_span/2] = 0
+  ch_g[abs(img)<zero_span/2] = img_max/2
+  ch_b[abs(img)<zero_span/2] = 0
+
+  return out
+
+# has to be pre transposed
+# it just suppose to match
+def tiles(img,shape,tiles_per_line=1,borders=True):
+
+  # shape is (n0,n1,n2,n3)
+  # n0*n1*n2*n3 = img.shape[1]
+  img_min = np.nanmin(img)
+  img_max = np.nanmax(img)
+
+  outer_color = [img_max,img_max,img_min]
+  outer_color = [img_max,img_max,img_max]
+
+  inner_color = [img_max/4,img_max/4,img_min]
+  inner_color = [img_min,img_min,img_min]
+  inner_color = [img_max,img_max,img_min]
+
+  group_h = shape[0]
+  group_w = shape[1]
+  group_size = group_h*group_w
+
+  tile_h = shape[2]
+  tile_w = shape[3]
+  tile_size = tile_h*tile_w
+
+  tpl = tiles_per_line
+
+  # main
+
+  tmp1 = []
+
+  for i in range(img.shape[0]):
+
+    if i%tpl==0:
+      tmp2 = []
+
+    tmp3 = []
+
+    for igh in range(group_h):
+
+      tmp4 = []
+
+      for igw in range(group_w):
+
+        si = (group_w*igh + igw + 0)*tile_size
+        ei = (group_w*igh + igw + 1)*tile_size
+
+        tile = img[i,si:ei]
+        tile = np.reshape(tile,(tile_h,tile_w,tile.shape[1]))
+
+        if borders:
+
+          if igw==group_w-1:
+
+            b_h_inner = [[inner_color]*(tile_w+0)]*(       1)
+            b_h_outer = [[outer_color]*(tile_w+0)]*(       1)
+            b_v_outer = [[outer_color]*(       1)]*(tile_h+1)
+
+            # outer hor
+            if igh==group_h-1:
+              tile = np.concatenate([tile,b_h_outer],axis=0)
+            # inner hor
+            else:
+              tile = np.concatenate([tile,b_h_inner],axis=0)
+            # outer vert
+            tile = np.concatenate([tile,b_v_outer],axis=1)
+
+          else:
+
+            b_v_inner = [[inner_color]*(       1)]*(tile_h+0)
+            b_h_inner = [[inner_color]*(tile_w+1)]*(       1)
+            b_h_outer = [[outer_color]*(tile_w+1)]*(       1)
+
+            # inner vert
+            tile = np.concatenate([tile,b_v_inner],axis=1)
+
+            # outer hor
+            if igh==group_h-1:
+              tile = np.concatenate([tile,b_h_outer],axis=0)
+            # inner hor
+            else:
+              tile = np.concatenate([tile,b_h_inner],axis=0)
+
+        tmp4.append(tile)
+
+      tmp3.append(np.concatenate(tmp4,axis=1))
+
+    tmp2.append(np.concatenate(tmp3,axis=0))
+
+    if i%tpl==(tpl-1):
+      tmp1.append(np.concatenate(tmp2,axis=1))
+
+  out = np.concatenate(tmp1,axis=0)
+  #out = img
+  return out
+
+if __name__=="__main__":
+  #
+  hw = hamming_window
+  #
+  image = np.array([[1*hw(i,512)*hw(j,512) for i in range(512)] for j in range(512)],np.float32)
+  zeros = np.zeros((512,512))
+
+  # 32x324
+
+  #image2 = np.zeros((32,324))
+  #rgb_img_0 = tiles(image2,(1,4,9,9),tiles_per_line=2,borders=True)
+
+  #image2 = np.zeros((32,144))
+  image2 = np.array([[1*hw(i,144)*hw(j,32) for i in range(144)] for j in range(32)],np.float32)
+  #image3 = coldmap(image2)
+  rgb_img_0 = tiles(coldmap(image2),(3,3,4,4),tiles_per_line=8,borders=True)
+
+  fig = plt.figure()
+  fig.suptitle("HaWi")
+  plt.imshow(rgb_img_0)
+
+  rgb_img = coldmap(image)
+
+  #print(rgb_img)
+
+  '''
+  for i in range(512):
+    for j in range(512):
+      if image[i,j]<0:
+        rgb_img[i,j,0] = 0
+        rgb_img[i,j,1] = 0
+        #rgb_img[i,j,2] = 255
+
+      if image[i,j]>0:
+        #rgb_img[i,j,0] = 255
+        rgb_img[i,j,1] = 0
+        rgb_img[i,j,2] = 0
+
+      if image[i,j]==0:
+        rgb_img[i,j,0] = 0
+        rgb_img[i,j,1] = 255
+        rgb_img[i,j,2] = 0
+  '''
+  print(rgb_img.shape)
+
+  fig = plt.figure()
+  fig.suptitle("HamWindow")
+  plt.imshow(rgb_img)
+  #plt.colorbar()
+  plt.show()
\ No newline at end of file