Unveiling The Mask of Position-Information Pattern Through the Mist of Image Features
This work addresses a specific issue in computer vision for researchers and practitioners, but it is incremental as it builds on prior studies of positional information in CNNs.
The paper tackles the problem of unreliable metrics for quantifying absolute position information encoded by padding in convolutional neural networks, proposing new metrics that measure and visualize this information more reliably than existing methods.
Recent studies show that paddings in convolutional neural networks encode absolute position information which can negatively affect the model performance for certain tasks. However, existing metrics for quantifying the strength of positional information remain unreliable and frequently lead to erroneous results. To address this issue, we propose novel metrics for measuring (and visualizing) the encoded positional information. We formally define the encoded information as PPP (Position-information Pattern from Padding) and conduct a series of experiments to study its properties as well as its formation. The proposed metrics measure the presence of positional information more reliably than the existing metrics based on PosENet and a test in F-Conv. We also demonstrate that for any extant (and proposed) padding schemes, PPP is primarily a learning artifact and is less dependent on the characteristics of the underlying padding schemes.