- Работаем с классификацией по брендам
- У некоторых брендов мало картинок.
- Отобрали только те бренды, у которых больше 100 картинок.
- Получилось 13 брендов для классификации.
- Разделили данные на train/val/test в пропорциях 60/20/20.
Feature-model_name | f1_macro | f1_micro | f1_weighted | accuracy |
---|---|---|---|---|
Baseline | 0.03 | 0.29 | 0.13 | 0.29 |
hog-log_reg | 0.72 | 0.76 | 0.76 | 0.76 |
hog-random_forest | 0.54 | 0.66 | 0.63 | 0.66 |
hog-decision_tree | 0.34 | 0.45 | 0.45 | 0.45 |
hog-svm | 0.78 | 0.81 | 0.81 | 0.81 |
hog-sgd | 0.72 | 0.76 | 0.75 | 0.76 |
hog-catboost | 0.67 | 0.73 | 0.71 | 0.73 |
sift-log_reg | 0.29 | 0.38 | 0.36 | 0.38 |
sift-random_forest | 0.18 | 0.38 | 0.29 | 0.38 |
sift-decision_tree | 0.17 | 0.24 | 0.24 | 0.24 |
sift-svm | 0.37 | 0.47 | 0.44 | 0.47 |
sift-sgd | 0.26 | 0.34 | 0.34 | 0.34 |
sift-catboost | 0.29 | 0.44 | 0.39 | 0.44 |
resnet152-log_reg | 0.71 | 0.73 | 0.73 | 0.73 |
resnet152-random_forest | 0.41 | 0.56 | 0.51 | 0.56 |
resnet152-decision_tree | 0.27 | 0.38 | 0.36 | 0.38 |
resnet152-svm | 0.76 | 0.76 | 0.76 | 0.76 |
resnet152-sgd | 0.72 | 0.74 | 0.74 | 0.74 |
resnet152-catboost | 0.64 | 0.7 | 0.68 | 0.7 |
Feature-model_name | f1_macro | f1_micro | f1_weighted | accuracy |
---|---|---|---|---|
Baseline | 0.03 | 0.29 | 0.13 | 0.29 |
hog-log_reg | 0.73 | 0.77 | 0.77 | 0.77 |
hog-random_forest | 0.51 | 0.66 | 0.62 | 0.66 |
hog-decision_tree | 0.34 | 0.44 | 0.44 | 0.44 |
hog-svm | 0.79 | 0.82 | 0.82 | 0.82 |
hog-sgd | 0.71 | 0.76 | 0.75 | 0.76 |
hog-catboost | 0.66 | 0.73 | 0.71 | 0.73 |
sift-log_reg | 0.28 | 0.39 | 0.37 | 0.39 |
sift-random_forest | 0.15 | 0.41 | 0.3 | 0.41 |
sift-decision_tree | 0.16 | 0.26 | 0.25 | 0.26 |
sift-svm | 0.36 | 0.46 | 0.43 | 0.46 |
sift-sgd | 0.25 | 0.38 | 0.36 | 0.38 |
sift-catboost | 0.32 | 0.48 | 0.43 | 0.48 |
resnet152-log_reg | 0.74 | 0.75 | 0.75 | 0.75 |
resnet152-random_forest | 0.4 | 0.57 | 0.51 | 0.57 |
resnet152-decision_tree | 0.28 | 0.36 | 0.36 | 0.36 |
resnet152-svm | 0.74 | 0.75 | 0.75 | 0.75 |
resnet152-sgd | 0.67 | 0.69 | 0.7 | 0.69 |
resnet152-catboost | 0.64 | 0.71 | 0.69 | 0.71 |
model_name | f1_macro | f1_micro | f1_weighted | accuracy |
---|---|---|---|---|
resnet152 | 0.85 | 0.88 | 0.88 | 0.88 |
vit | 0.83 | 0.87 | 0.87 | 0.87 |