About this classifier

THIS IS A CLASSIFIER WHICH I BUILT AS PART OF A PROOF OF CONCEPT FOR THE POLICE OFFICE IN TLAXCALA, MEXICO.

THE OBJECTIVE IS TO CLASSIFY TATTOO IMAGES INTO THREE TYPES OF IMAGES: WOMEN, SKULLS AND MONSTERS. I TRAINED THIS MODEL USING 300 TATTOO IMAGES: 100 IMAGES CORRESPONDING TO SKULLS, 100 CORRESPONDING TO WOMEN AND 100 CORRESPONDING TO MONSTERS. I SPLIT THEM INTO 70% FOR TRAINING AND 30% FOR TESTING

About the neural network used

THIS CLASSIFIER IS BASED ON A MULTILAYER PERCEPTRON NEURAL NETWORK WHICH I BUILT USING TENSORFLOW. IT HAS ONLY ONE HIDDEN LAYER WITH 1000 NEURONS, THE LEARNING RATE IS EQUAL TO 0.01 AND I SET 1000 EPOCHS. THE ACCURACY SCORE ON THE TESTING IS 45%. THIS WAS THE HIGHEST SCORE USING 1 HIDDEN LAYER AND MANUAL GRID SEARCH.

THE SOURCE CODE IS AVAILABLE HERE: https://github.com/antonioACR1/classifier/blob/master/classifier.ipynb

Final comments

TO SEE HOW IT WORKS, SIMPLY UPLOAD AN IMAGE AND CLICK THE BUTTON. THEN WAIT A FEW SECONDS. IT WILL DISPLAY THE PREDICTED CLASSIFICATION: 'MUJER' (WOMAN), 'MONSTRUO' (MONSTER) OR 'CALAVERA' (SKULL). NOTE HOWEVER THAT IT MIGHT SHOW AN ERROR DEPENDING ON THE FILE YOU'RE TRYING TO UPLOAD.

PLEASE NOTE THAT THIS IS JUST A PROOF OF CONCEPT! I WORKED ON THIS PROOF OF CONCEPT FOR A COUPLE OF WEEKS AND I SPENT MOST OF MY TIME LEARNING LEARNING HOW TO BUILD A SIMPLE NEURAL NETWORK WITH TENSORFLOW RATHER THAN TRYING TO IMPROVE THE ACCURACY.