DarkCoder999/Image-Classification

Developed a Transformer-based model using ViT that can accurately classify natural scene images into one of the six predefined categories: buildings, forest, glacier, mountain, sea, and street

Developed a Transformer-based model using ViT that can accurately classify natural scene images into one of the six predefined categories: buildings, forest, glacier, mountain, sea, and street

Dataset

Model name: vit_tiny_patch16_224

category_list = {"mountain":0,"street":1,"buildings":2,"sea":3,"forest":4,"glacier":5}
TRAIN LOSS : 0.00484765412424284
TRAIN PREC : [0.99799679 0.9978903 0.99907961 0.99955733 1. 0.99790882]
RECALL : [0.99759712 0.99915505 0.99770221 0.99955733 0.99955674 0.99874424]
F1 : [0.99779692 0.99852227 0.99839043 0.99955733 0.99977832 0.99832636]
ACC : 99.87098623853211

VALIDATION LOSS : 0.3094542117396486
VALIDATION PREC : [0.93333333 0.92307692 0.875 1. 1. 0.88235294]
RECALL : [0.93333333 0.8 0.93333333 0.93333333 1. 1. ]
F1 : [0.93333333 0.85714286 0.90322581 0.96551724 1. 0.9375 ]
ACC : 92.5

TEST ACC : 93.67