Abonneren om te downloaden
Handwritten Digit Recognizer V2.1

BE CAREFUL INSTALLING THIS VERSION, BETTER INSTALL THE V2 VERSION, V2.1 VERSION WEIGHTS 382 MB IT MIGHT CRASH YOUR GAME OR CORRUPT SAVE FILES, DOWNLOAD AT YOUR OWN RISK

Handwritten Digit Recognizer V2.1

This version has more models

Trained in Python using MNIST dataset
Made in 2 Days from zero without any libraries or any wifi modules
44x44 px screen 4x4 px size

Changes:
Models now split into different files to optimize performance and allow adding multiple models without making game lag on start or when viewing code
Added more diffrent models

New models:
1. PowerfulModel V2
Accuracy: 98.11%
1936 - Input Layer
128 - Hidden Layer 1
128 - Hidden Layer 2
128 - Hidden Layer 3
10 - Output Layer
2. VeryLongModel
Accuracy: 98.22%
1936 - Input Layer
256 - Hidden Layer 1
128 - Hidden Layer 2
64 - Hidden Layer 3
32 - Hidden Layer 4
16 - Hidden Layer 5
10 - Output Layer
3. FastModel
Accuracy: 98.7%
1936 - Input Layer
256 - Hidden Layer 1
256 - Hidden Layer 2
256 - Hidden Layer 3
10 - Output Layer
4. LongModel
Accuracy: 98.1%
1936 - Input Layer
128 - Hidden Layer 1
128 - Hidden Layer 2
128 - Hidden Layer 3
128 - Hidden Layer 4
128 - Hidden Layer 5
10 - Output Layer
5. SlowModel
Accuracy: 98%
1936 - Input Layer
512 - Hidden Layer 1
256 - Hidden Layer 2
128 - Hidden Layer 3
10 - Output Layer
6. BigModel
Accuracy: 98.5%
1936 - Input Layer
1936 - Hidden Layer 1
10 - Output Layer
7. LARGEModel
Accuracy: 98.35%
1936 - Input Layer
1936 - Hidden Layer 1
1936 - Hidden Layer 2
10 - Output Layer