Project of Cal CS294-082 Fall 2021
Create a data folder and place training and test data in torch.ImageFolder style.
I directly use the image pixel as random forest input, another three types of feature are also applied in notebook.
Run the Random-Forest-Image-Classification-using-Python.ipynb
Result
training acc: 0.9985022118499425
training acc: 0.4856505990526609
Model's confusion matrix on training data
[[3988 0 1 0 1 4 1]
[ 0 435 0 0 0 0 1]
[ 3 1 4080 0 3 3 7]
[ 1 0 0 7213 0 0 1]
[ 2 0 0 0 4962 1 0]
[ 5 0 4 0 0 4821 0]
[ 1 0 3 0 0 0 3167]]
Model's confusion matrix on test data
[[ 199 0 53 355 142 163 46]
[ 6 32 5 43 10 10 5]
[ 43 0 281 293 158 162 87]
[ 20 0 38 1423 102 146 45]
[ 20 0 35 430 527 193 28]
[ 36 0 62 405 223 500 21]
[ 14 0 45 135 61 52 524]]