ashhadahsan's picture
Update README.md
713c2da verified
metadata
language:
  - en
  - de
base_model:
  - facebook/sam-vit-base

Training Loss UFPR SAM

Comparison with baseline

Finetuning metrics

  • RANK LOSS : 0.052654343576998344
  • RANK ACCURACY : 0.9997852160627786
  • RANK PRECISION : 0.9420321321139293
  • RANK RECALL : 0.9461407410472382

Parameters on which the model was trained

DATASET:
  TRAIN_PATH: "./ufpr_dataset/training"
  TEST_PATH: "./ufpr_dataset/testing"
  IMAGE_FORMAT: ".png"

SAM:
  CHECKPOINT: "./sam_vit_b_01ec64.pth"
  RANK: 512

TRAIN:
  BATCH_SIZE: 1
  NUM_EPOCHS: 30
  LEARNING_RATE: 0.0001
  PATIENCE: 10