-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
45 lines (35 loc) · 1.96 KB
/
main.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
import numpy as np
from preprocessDataset import preprocessDataset
from distributeDataset import distributeDatset
from trainModel import trainModel
from testModel import testModel
if __name__ == "__main__":
# Training Variables
MARGIN = 25
RESNET_SIZE = "ResNet34" # "ResNet34" or "ResNet50" or "ResNet101" or "ResNet152"
MIN_AGE = 13 # Inclusive
MAX_AGE = 116 # Inclusive
M_array = [2, 4, 8, 16, 32, 64] # Number of different bin configurations
L_array = [8, 16, 32, 64] # Number of bins in each configuration
SOCIAL_MEDIA_SEGMENTS = np.array([25, 35, 50]) # (MIN_AGE,24), (25,34), (35,49), (50,MAX_AGE)
# Raw Images Paths
DATASETS_PATH = "../Datasets/"
# Preprocessed Images Paths
PREPROCESSED_FOLDER_PATH = DATASETS_PATH + "UTKFace-Preprocessed-" + str(MARGIN) + "/"
PREPROCESSED_IMAGES_PATH = PREPROCESSED_FOLDER_PATH + "Images/"
PREPROCESSED_CSV_PATH = PREPROCESSED_FOLDER_PATH + "CSVs/"
# preprocessDataset(MARGIN, DATASETS_PATH, PREPROCESSED_FOLDER_PATH, PREPROCESSED_IMAGES_PATH, PREPROCESSED_CSV_PATH)
# distributeDatset(PREPROCESSED_CSV_PATH, MIN_AGE, MAX_AGE, SOCIAL_MEDIA_SEGMENTS)
best_age_mae, best_ML = -1, (0, 0)
for M in M_array:
for L in L_array:
# Output Path
OUTPUT_FOLDER_NAME = "LabelDiversity-" + RESNET_SIZE + "-" + str(M) + "X" + str(L)
OUT_PATH = "../TrainedModels/" + OUTPUT_FOLDER_NAME + "/"
validLoss = trainModel(PREPROCESSED_IMAGES_PATH, PREPROCESSED_CSV_PATH, OUT_PATH, RESNET_SIZE, MIN_AGE, MAX_AGE, M, L)
age_mae = testModel(PREPROCESSED_IMAGES_PATH, PREPROCESSED_CSV_PATH, OUT_PATH, SOCIAL_MEDIA_SEGMENTS, RESNET_SIZE, MIN_AGE, MAX_AGE, M, L)
if age_mae < best_age_mae or best_age_mae == -1:
best_age_mae = age_mae
best_ML = (M, L)
print("Current ML:\t", (M, L), "\tCurrent MAE:\t", age_mae)
print("Best ML:\t", best_ML, "\tBest MAE:\t", best_age_mae)