StableTTA: Training-Free Test-Time Adaptation that Improves Model Accuracy on ImageNet1K to 96%

1. 1Identify the conflict in aggregation strategies that negatively impacts prediction stability
2. 2Propose a training-free method to improve aggregation stability and efficiency
3. 3Apply StableTTA to existing models to improve their performance on tasks like image classification
4. 4Evaluate the effectiveness of StableTTA using empirical results on benchmark datasets like ImageNet-1K

Machine learning researchers and engineers can benefit from StableTTA to improve model performance without requiring additional training data or computational resources. This can be particularly useful for teams working on computer vision tasks

💡 StableTTA improves model accuracy without requiring additional training data or computational resources

🚀 StableTTA: training-free test-time adaptation boosts ImageNet1K accuracy to 96%