updates from colab [fix model, reduce augmentation]

aug.py

@@ -4,6 +4,7 @@ from PIL import Image
 from os import listdir as ls
 from sys import argv, exit
 from tensorflow.keras.utils import array_to_img as img
+from tensorflow.keras import layers as kl
 
 if len(argv) != 3:
     exit(1)
@@ -14,11 +15,11 @@ for v, fn in enumerate(files):
     i = np.array(Image.open(f"{argv[1]}/{fn}").convert('RGB').resize((300,300))) / 255.0
     print(f"Processing {v+1:03d}/{len(files)}: {fn}")
 
-    for x1 in range(4):
+    for x1 in range(3):
         i1 = tf.image.random_brightness(i, 0.6)
-        for x2 in range(4):
+        for x2 in range(3):
             i2 = tf.image.random_saturation(i1, 0.1, 2.0)
-            for x3 in range(4):
+            for x3 in range(3):
                 i3 = tf.image.random_contrast(i2, 0.2, 1.9)
                 img(i3).save(f"{argv[2]}/{fn.rsplit('.', 1)[0]}_{x1}_{x2}_{x3}_1.png")
 

m.py

@@ -5,9 +5,10 @@ from tensorflow import keras as k
 from tensorflow.keras import layers as kl
 from tensorflow.keras import models as km
 from tensorflow.keras import optimizers as ko
+from tensorflow.keras import callbacks as kc
 
-CONV_SIZE = 128
-MPPT_SIZE = 1024
+CONV_SIZE = 96
+MPPT_SIZE = 256
 DRPT_RATE = 0.3
 TOUT_AMNT = 2
 
@@ -23,7 +24,8 @@ def max(i, s):
     return kl.MaxPooling2D(s, padding = 'same', strides = (1, 1))(i)
 
 def generate_start(i):
-    r1 = conv(i, CONV_SIZE, (3, 3))
+    s = kl.Rescaling(1./255.)(i)
+    r1 = conv(s, CONV_SIZE, (3, 3))
     r2 = conv(r1, CONV_SIZE, (3, 3))
     r3 = conv(r2, CONV_SIZE, (3, 3))
 
@@ -133,19 +135,19 @@ gi = kl.Input((300, 300, 3))
 
 ds = generate_start(gi)
 
-for _ in range(1):
+for _ in range(3):
     ds = generate_type_a(ds)
 
 ds = generate_ab_bridge(ds)
 
-for _ in range(1):
+for _ in range(4):
     ds = generate_type_b(ds)
 
 uo = generate_aux(ds)
 
 go = generate_bc_bridge(ds)
 
-for _ in range(1):
+for _ in range(2):
     go = generate_type_c(go)
 
 go = generate_finish(go)
@@ -153,6 +155,6 @@ go = generate_finish(go)
 mod = k.Model(inputs = gi, outputs = go)
 
 mod.summary()
-mod.compile(optimizer = ko.Lion(learning_rate = 0.001),
+mod.compile(optimizer = ko.Lion(learning_rate = 0.0001),
             loss = 'categorical_crossentropy',
             metrics = ['accuracy'])