CubeBit/StreetsignRecognition
1
0
Fork 0
mirror of https://github.com/3x4byte/StreetsignRecognition.git synced 2026-03-21 21:25:06 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

READY

Browse Source
This commit is contained in:
Denys Seredenko
2024-11-26 16:52:53 +01:00
parent 61ab3810e2
commit f09861a01c
4 changed files with 48 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
src/all_feature_vectors.csv
View File

@@ -1,5 +1,4 @@
brightness;filename;concept;vector
+1;X-10Y-10.png;Concept.LINKS_ABBIEGEN;[0, 58, 0, 40, 0, 2, 0, 18, 0, 79, 0, 3, 0, 64, 0, 33, 0, 3, 0, 79, 0, 18, 0, 3, 0, 21, 0, 76, 0, 3, 0, 69, 0, 28, 0, 3, 0, 39, 0, 60, 0, 1, 0, 68, 0, 31, 0, 1, 0, 77, 0, 20, 0, 3, 0, 79, 0, 18, 0, 3, 0, 100, 0, 0, 0, 0, 0, 62, 0, 37, 0, 1, 0, 73, 0, 23, 0, 4, 0, 20, 0, 76, 0, 4, 0, 77, 0, 20, 0, 3, 0, 62, 0, 34, 0, 4, 0, 15, 0, 82, 0, 3, 9, 33, 3]
+1;X-10Y-20.png;Concept.LINKS_ABBIEGEN;[0, 58, 0, 40, 0, 2, 0, 17, 0, 81, 0, 2, 0, 64, 0, 33, 0, 3, 0, 80, 0, 17, 0, 3, 0, 22, 0, 75, 0, 3, 0, 65, 0, 32, 0, 3, 0, 41, 0, 57, 0, 2, 0, 71, 0, 28, 0, 1, 0, 77, 0, 20, 0, 3, 0, 81, 0, 16, 0, 3, 0, 100, 0, 0, 0, 0, 0, 61, 0, 38, 0, 1, 0, 70, 0, 26, 0, 4, 0, 23, 0, 73, 0, 4, 0, 78, 0, 18, 0, 4, 0, 59, 0, 37, 0, 4, 0, 12, 0, 85, 0, 3, 9, 32, 3]
+1;X-10Y-30.png;Concept.LINKS_ABBIEGEN;[0, 57, 0, 40, 0, 3, 0, 15, 0, 83, 0, 2, 0, 63, 0, 34, 0, 3, 0, 81, 0, 16, 0, 3, 0, 23, 0, 73, 0, 4, 0, 61, 0, 36, 0, 3, 0, 43, 0, 55, 0, 2, 0, 75, 0, 24, 0, 1, 0, 76, 0, 20, 0, 4, 0, 83, 0, 14, 0, 3, 0, 99, 0, 0, 0, 1, 0, 61, 0, 38, 0, 1, 0, 66, 0, 30, 0, 4, 0, 27, 0, 70, 0, 3, 0, 78, 0, 18, 0, 4, 0, 57, 0, 39, 0, 4, 0, 9, 0, 88, 0, 3, 7, 31, 2]
@@ -43336,4 +43335,4 @@ brightness;filename;concept;vector
0;X80Y50.png;Concept.VORFAHRT_GEWAEHREN;[43, 0, 0, 0, 0, 57, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 80, 0, 0, 0, 0, 20, 100, 0, 0, 0, 0, 0, 50, 0, 0, 0, 0, 50, 40, 0, 0, 0, 0, 60, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 90, 0, 0, 0, 0, 10, 60, 0, 0, 0, 0, 40, 50, 0, 0, 0, 0, 50, 20, 0, 0, 0, 0, 80, 0, 0, 0, 0, 0, 100, 9, 0, 1]
0;X80Y60.png;Concept.VORFAHRT_GEWAEHREN;[38, 0, 0, 0, 0, 62, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 20, 0, 0, 0, 0, 80, 90, 0, 0, 0, 0, 10, 60, 0, 0, 0, 0, 40, 20, 0, 0, 0, 0, 80, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 90, 0, 0, 0, 0, 10, 60, 0, 0, 0, 0, 40, 50, 0, 0, 0, 0, 50, 20, 0, 0, 0, 0, 80, 0, 0, 0, 0, 0, 100, 4, 0, 1]
0;X80Y70.png;Concept.VORFAHRT_GEWAEHREN;[34, 0, 0, 0, 0, 66, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 40, 0, 0, 0, 0, 60, 80, 0, 0, 0, 0, 20, 25, 0, 0, 0, 0, 75, 80, 0, 0, 0, 0, 20, 100, 0, 0, 0, 0, 0, 90, 0, 0, 0, 0, 10, 50, 0, 0, 0, 0, 50, 60, 0, 0, 0, 0, 40, 10, 0, 0, 0, 0, 90, 0, 0, 0, 0, 0, 100, 3, 0, 1]
0;X80Y80.png;Concept.VORFAHRT_GEWAEHREN;[29, 0, 0, 0, 0, 71, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 20, 0, 0, 0, 0, 80, 70, 0, 0, 0, 0, 30, 12, 0, 0, 0, 0, 88, 70, 0, 0, 0, 0, 30, 100, 0, 0, 0, 0, 0, 80, 0, 0, 0, 0, 20, 50, 0, 12, 0, 0, 38, 50, 0, 0, 0, 0, 50, 10, 0, 0, 0, 0, 90, 0, 0, 0, 0, 0, 100, 2, 0, 0]
0;X80Y80.png;Concept.VORFAHRT_GEWAEHREN;[29, 0, 0, 0, 0, 71, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 20, 0, 0, 0, 0, 80, 70, 0, 0, 0, 0, 30, 12, 0, 0, 0, 0, 88, 70, 0, 0, 0, 0, 30, 100, 0, 0, 0, 0, 0, 80, 0, 0, 0, 0, 20, 50, 0, 12, 0, 0, 38, 50, 0, 0, 0, 0, 50, 10, 0, 0, 0, 0, 90, 0, 0, 0, 0, 0, 100, 2, 0, 0]
Can't render this file because it is too large.

4
src/classes/neural/network.py
View File

@@ -33,11 +33,11 @@ class NeuralNetwork:
return [interpritated, True]
# takes feature vectors to train and generates weights
def learn(self, training_set: list[FeatureVector]):
def learn(self, training_set: list[FeatureVector], epochs = 10):
correct_amount = 0
false_amount = 0
for i in range(len(training_set)*100):
for i in range(len(training_set)*epochs):
fv_to_train = training_set[i%len(training_set)] # random.choice(training_set)
classified_concept, correct = self.classify(fv_to_train)

85
src/main.py
View File

@@ -2,6 +2,7 @@ import os
import ast
import csv
import numpy as np
import random
import multiprocessing
from classes.concept import Concept
@@ -59,56 +60,60 @@ def decision_tree(training = os.path.abspath(os.path.join(__file__, "..", "train
# print(f"classified {cnt_correct}/{cnt_data} correctly ({round(cnt_correct/cnt_data*100, 3)}%)")
return [cnt_correct, cnt_data - cnt_correct, cnt_data] # richtig, falsch, amount of testing data
def neural_network(training = os.path.abspath(os.path.join(__file__, "..", "training.csv")), testing = os.path.abspath(os.path.join(__file__, "..", "testing.csv")), num_training_vectors = 100):
def neural_network(training_set, testing_set, epochs = 10):
from classes.neural.network import NeuralNetwork
from classes.neural.neuron import Neuron
training_data = []
with open(training, 'r') as csv_file:
reader = csv.reader(csv_file, delimiter=";")
next(reader)
i = 0
for row in reader:
if i > num_training_vectors:
break
i += 1
float_list = ast.literal_eval(row[3])
fv = FeatureVector(concept = Concept.identify_by_str(row[2].split(".")[1]), features_list = [element * 0.01 for element in float_list], loaded = True)
training_data.append(fv)
while len(training_data) < num_training_vectors:
training_data += training_data
training_data = training_data[:num_training_vectors+1]
nn = NeuralNetwork(neurons=None)
nn.learn(training_set, epochs)
nn = NeuralNetwork()
nn.learn(training_data)
cnt_data = 0
cnt_correct = 0
with open(testing, 'r') as csv_file:
reader = csv.reader(csv_file, delimiter=";")
next(reader)
for row in reader:
cnt_data += 1
fv = FeatureVector(concept = Concept.identify_by_str(row[2].split(".")[1]), features_list = ast.literal_eval(row[3]), loaded = True)
classified_concept, _ = nn.classify(fv)
cnt_correct += 1 if fv.concept == classified_concept else 0
# print(f"{fv.concept} was classified as {classified_concept}")
for testing_fv in testing_set:
classified_concept, _ = nn.classify(testing_fv)
cnt_correct += 1 if testing_fv.concept == classified_concept else 0
# print(f"{fv.concept} was classified as {classified_concept}")
#print(f"classified {cnt_correct}/{cnt_data} correctly ({round(cnt_correct/cnt_data*100, 3)}%)")
return [cnt_correct, cnt_data - cnt_correct, cnt_data] # richtig, falsch, amount of testing data
return [cnt_correct, len(testing_set) - cnt_correct, len(testing_set)] # richtig, falsch, amount of testing data
def repeat_n_times(n: int, num_training_vectors: int):
def repeat_n_times(n: int, num_training_vectors: int, epochs: int):
cnt_correct = 0
cnt_data = 0
for _ in range(n):
cnt_correct = 0
cnt_data = 0
result = neural_network(num_training_vectors=num_training_vectors)
result = neural_network(num_training_vectors=num_training_vectors, epochs=epochs)
cnt_correct += result[0]
cnt_data += result[2]
print(f"{num_training_vectors}: classified {cnt_correct}/{cnt_data} correctly ({round(cnt_correct/cnt_data*100, 3)}%)")
print(f"{epochs}: classified {cnt_correct}/{cnt_data} correctly ({round(cnt_correct/cnt_data*100, 3)}%)")
def load_fv(path_to_training_data = os.path.abspath(os.path.join(__file__, "..", "all_feature_vectors.csv"))):
training_data = []
with open(path_to_training_data, 'r') as csv_file:
reader = csv.reader(csv_file, delimiter=";")
next(reader)
for row in reader:
fv_float = ast.literal_eval(row[3])
fv = FeatureVector(concept = Concept.identify_by_str(row[2].split(".")[1]), features_list = [element * 0.01 for element in fv_float], loaded = True)
training_data.append(fv)
return training_data
if __name__ == "__main__":
for num_training_vectors in range(100, 1000, 100):
multiprocessing.Process(target=repeat_n_times, args=(100, num_training_vectors, )).start()
all_vectors = load_fv()
for input in range(40000, len(all_vectors), 400):
inp_res = []
for epoch in range(10, 1000, 10):
l = []
for test in range(0, 1):
training_set = random.sample(all_vectors, input)
testing_set = list(set(all_vectors)-set(training_set))
# print(f"Start for {input} trainingsdata and test number {test} with epoch {epoch}")
correct, false, ammount = neural_network(training_set, testing_set, epoch)
print(f"Epoch {epoch}: correct_percentage={(correct/ammount)*100}%")
l.append(false/ammount)
inp_res.append(np.mean(l))

1
src/training.csv
View File

@@ -2167,7 +2167,6 @@ brightness;filename;concept;vector
+1;X-10Y-50.png;Concept.VORFAHRT_STRASSE;[0, 0, 15, 77, 0, 8, 0, 0, 0, 99, 0, 1, 0, 0, 3, 86, 0, 11, 0, 0, 1, 85, 0, 14, 0, 0, 0, 99, 0, 1, 0, 0, 11, 79, 0, 10, 0, 0, 85, 12, 0, 3, 0, 0, 35, 60, 0, 5, 0, 0, 1, 82, 0, 17, 0, 1, 10, 78, 0, 11, 0, 0, 79, 17, 0, 4, 0, 0, 16, 74, 0, 10, 0, 0, 1, 88, 0, 11, 0, 0, 0, 92, 0, 8, 0, 0, 1, 86, 0, 13, 0, 0, 0, 94, 0, 6, 0, 0, 0, 100, 0, 0, 10, 38, 1]
+2;X50Y50.png;Concept.STOP;[43, 0, 0, 0, 0, 57, 0, 0, 0, 0, 0, 100, 8, 0, 0, 0, 0, 92, 65, 0, 0, 0, 0, 35, 73, 0, 0, 0, 0, 27, 11, 0, 0, 0, 0, 89, 65, 0, 0, 0, 0, 35, 70, 0, 0, 0, 0, 30, 64, 0, 0, 0, 0, 36, 48, 0, 0, 0, 0, 52, 59, 0, 0, 0, 0, 41, 57, 0, 0, 0, 0, 43, 10, 0, 0, 0, 0, 90, 72, 0, 0, 0, 0, 28, 74, 0, 0, 0, 0, 26, 15, 0, 0, 0, 0, 85, 0, 0, 0, 0, 0, 100, 9, 16, 5]
-1;X-80Y10.png;Concept.STOP;[49, 0, 0, 49, 0, 2, 22, 0, 0, 76, 0, 2, 71, 0, 0, 27, 0, 2, 54, 0, 0, 44, 0, 2, 6, 0, 0, 93, 0, 1, 48, 0, 0, 49, 0, 3, 64, 0, 0, 34, 0, 2, 47, 0, 0, 50, 0, 3, 44, 0, 0, 54, 0, 2, 44, 0, 0, 54, 0, 2, 98, 0, 0, 1, 0, 1, 98, 0, 0, 1, 0, 1, 62, 0, 0, 36, 0, 2, 0, 0, 0, 98, 0, 2, 49, 0, 0, 48, 0, 3, 55, 0, 0, 42, 0, 3, 17, 0, 0, 80, 0, 3, 19, 13, 56]
+1;X60Y-80.png;Concept.VORFAHRT_STRASSE;[0, 0, 3, 94, 0, 3, 0, 0, 0, 93, 0, 7, 0, 0, 0, 98, 0, 2, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 99, 0, 1, 0, 0, 21, 72, 0, 7, 0, 0, 2, 94, 0, 4, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 1, 97, 0, 2, 0, 0, 28, 66, 0, 6, 0, 0, 0, 97, 0, 3, 0, 0, 0, 100, 0, 0, 0, 0, 0, 100, 0, 0, 0, 0, 0, 99, 0, 1, 0, 0, 1, 92, 0, 7, 16, 19, 3]
-2;X10Y60.png;Concept.VORFAHRT_GEWAEHREN;[24, 0, 0, 74, 0, 2, 59, 0, 0, 35, 1, 5, 43, 0, 0, 53, 0, 4, 34, 0, 0, 63, 0, 3, 11, 0, 0, 87, 0, 2, 38, 0, 0, 58, 0, 4, 16, 0, 0, 83, 0, 1, 3, 0, 0, 95, 0, 2, 40, 0, 0, 58, 0, 2, 4, 0, 0, 93, 0, 3, 45, 0, 0, 53, 0, 2, 28, 0, 0, 70, 0, 2, 3, 0, 0, 95, 0, 2, 0, 0, 0, 99, 0, 1, 43, 0, 0, 52, 0, 5, 11, 0, 0, 87, 0, 2, 0, 0, 0, 99, 0, 1, 8, 23, 5]
0;X-40Y10.png;Concept.VORFAHRT_STRASSE;[0, 0, 15, 77, 0, 8, 0, 0, 0, 97, 0, 3, 0, 0, 7, 80, 0, 13, 0, 0, 9, 81, 0, 10, 0, 0, 0, 98, 0, 2, 0, 0, 0, 85, 0, 15, 0, 0, 75, 20, 0, 5, 0, 0, 87, 10, 0, 3, 0, 0, 4, 83, 0, 13, 0, 0, 0, 91, 0, 9, 0, 0, 16, 74, 0, 10, 0, 0, 41, 53, 0, 6, 0, 0, 0, 86, 0, 14, 0, 0, 0, 100, 0, 0, 0, 0, 0, 90, 0, 10, 0, 0, 0, 80, 6, 14, 0, 0, 0, 99, 0, 1, 8, 43, 1]
Can't render this file because it is too large.