cpoisson's picture
Add efficientnet_custom_v2_2 trained on 1000 images per class and augmentation of unbalanced classes
38e2d68
Classification Report
================================================================================
precision recall f1-score support
battery 1.0000 0.9928 0.9964 276
car_battery 0.9959 1.0000 0.9979 243
cardboard 0.9062 0.9170 0.9116 253
food_organics 0.9436 0.9366 0.9401 268
glass 0.9685 0.9179 0.9425 268
light_bulb 1.0000 0.9883 0.9941 256
metal 0.8826 0.8859 0.8843 263
mirror 0.9957 1.0000 0.9978 229
miscellaneous_trash 0.9121 0.8755 0.8934 249
neon 0.9784 0.9956 0.9869 227
paper 0.8996 0.8630 0.8809 270
pharmacy 0.9881 0.9921 0.9901 252
plastic 0.8000 0.8681 0.8327 235
printer_cartridge 0.9668 0.9957 0.9811 234
textile_trash 0.9277 0.9083 0.9179 240
tire 0.9763 0.9960 0.9860 248
vegetation 0.9810 0.9961 0.9885 259
wood 1.0000 1.0000 1.0000 230
accuracy 0.9507 4500
macro avg 0.9512 0.9516 0.9512 4500
weighted avg 0.9511 0.9507 0.9507 4500