Explain To Me: Salience-Based Explainability for Synthetic Face Detection Models
This work addresses the need for scalable explainability in human-machine pairing setups and model selection, though it is incremental in extending salience-based analysis.
The paper tackles the challenge of explaining complex model decisions at scale by proposing five novel methods that leverage salience maps to analyze model behavior, and evaluates them on synthetic face detection, showing interpretable differences between traditionally trained models and those guided by human salience.
The performance of convolutional neural networks has continued to improve over the last decade. At the same time, as model complexity grows, it becomes increasingly more difficult to explain model decisions. Such explanations may be of critical importance for reliable operation of human-machine pairing setups, or for model selection when the "best" model among many equally-accurate models must be established. Saliency maps represent one popular way of explaining model decisions by highlighting image regions models deem important when making a prediction. However, examining salience maps at scale is not practical. In this paper, we propose five novel methods of leveraging model salience to explain a model behavior at scale. These methods ask: (a) what is the average entropy for a model's salience maps, (b) how does model salience change when fed out-of-set samples, (c) how closely does model salience follow geometrical transformations, (d) what is the stability of model salience across independent training runs, and (e) how does model salience react to salience-guided image degradations. To assess the proposed measures on a concrete and topical problem, we conducted a series of experiments for the task of synthetic face detection with two types of models: those trained traditionally with cross-entropy loss, and those guided by human salience when training to increase model generalizability. These two types of models are characterized by different, interpretable properties of their salience maps, which allows for the evaluation of the correctness of the proposed measures. We offer source codes for each measure along with this paper.