mirror of
https://github.com/VitaAetaerna/PetDetectionPy.git
synced 2024-09-20 03:00:58 +02:00
37 lines
1.1 KiB
Plaintext
37 lines
1.1 KiB
Plaintext
!!!! Made by Vita_Aeterna and Google researches. !!!!
|
|
©Used DataSet "https://www.microsoft.com/en-us/download/details.aspx?id=54765"
|
|
|
|
|
|
|
|
|
|
Simple tensorflow Neural Network with Sigmoid and Binary_Crossentropy, learning with and about accuracy and loss.
|
|
All of the Pictures are converted to (50,50) sized Pictures and also GrayScaled (To improve the learning process of the Neural Network).
|
|
At first i tested multiple Versions(different Dense layers, Conv layers and Layer size) of The Neural Network, The one with the best Accuracy and smallest Loss is used to predict the Animals.
|
|
I noticed that if u give an invalid picture like a Man or a Woman the Neural Network will always give the Ouput 0!
|
|
|
|
|
|
|
|
|
|
|
|
As Input a Picture of a Dog or a Cat is given (Need to change Name of your Picture in "EndProduct" script, default Name is "dog.jpg")
|
|
Note that the Picture has to be in png or jpg Format.
|
|
Path of Cat and Dog Pictures (DataSet) in "ImageUsing" need to be changed too! (DATA_DIR)
|
|
|
|
|
|
|
|
As Output, 0 or 1 is given.
|
|
0=Dog;
|
|
1=Cat;
|
|
|
|
All packages (pip) used in this Project:
|
|
tensorflow / tensorflow-gpu
|
|
pickle
|
|
numpy
|
|
random
|
|
os
|
|
matplotlib
|
|
tqdm
|
|
cv2
|
|
time
|
|
|