Experimental performance records

Params

Pre-trained weights: R50+ViT-B/16
Input shape: (224, 224, 3)
Encoder trainable: False
Batch size: 24
Epochs: 1,290

Model

Input Image (224, 224, 3) → Vision Transformer (224, 224, 32) → Conv1x1 (224, 224, 1)

Total params: 100,840,640
Trainable params: 7,389,312
Non-trainable params: 93,451,328

Dataset

• Medial Images for Gastric Cancer Diagnosis (Korean)

Preprocessing

• Preprocessors

1. Remove malformed images from entire dataset.
2. Fill missing polygons from label data using CVAT. (Thanks to @kdh93)
3. Exports its ROI, binary masks and transformed original images.

Augmentation

Used tf.data.Dataset to boost training performance

• Scaling (1.0 / 255)
• Random Flip Left-Right
• Random Flip Up-Down
• Random Crop
• Random Brightness (-0.2 ~ +0.2)
• Gaussian Noise (mean = 0, stddev = 0.05)

Compile options

Loss

• BCE * 0.5 + Dice * 0.5

Metrics

• Binary IoU
  • Number of classes = 2 (default)
  • Threshold = 0.5

Optimizer

• SGD
  • Momentum = 0.9
  • Learning Rate Scheduler
    • Cosine Annealing Warmup Restarts
      • First cycle steps = 100
      • Initial learning rate = 1e-3
      • First decay steps = 300
      • t_mul = 1.0
      • m_mul = 1.0 (default)
      • alpha = 0.0 (default)

Cosine Annealing Warmup Restarts:

Results

Evaluation