dima806's picture
Update README.md
0e67933 verified
metadata
license: apache-2.0
metrics:
  - accuracy
base_model:
  - google/vit-base-patch16-224-in21k

Achieved 78% weighted accuracy for classification between 3 common galaxy types (S, E, Sb).

See my Kaggle notebook for more details.

image/png

Classification report:

              precision    recall  f1-score   support

           E     0.7656    0.8848    0.8209     13592
           S     0.7526    0.6685    0.7081     13591
          SB     0.8262    0.7900    0.8077     13591

    accuracy                         0.7811     40774
   macro avg     0.7815    0.7811    0.7789     40774
weighted avg     0.7815    0.7811    0.7789     40774