CYBORG: Blending Human Saliency Into the Loss Improves Deep Learning

Abstract

Can deep learning models achieve greater generalization if their training is guided by reference to human perceptual abilities? And how can we implement this in a practical manner? This paper proposes a first-ever training strategy to ConveY Brain Oversight to Raise Generalization (CY-BORG). This new training approach incorporates humanannotated saliency maps into a CYBORG loss function that guides the model towards learning features from image regions that humans find salient when solving a given visual task. The Class Activation Mapping (CAM) mechanism is used to probe the model's current saliency in each training batch, juxtapose model saliency with human saliency, and penalize the model for large differences. Results on the task of synthetic face detection show that the CYBORG loss leads to significant improvement in performance on unseen samples consisting of face images generated from six Generative Adversarial Networks (GANs) across multiple classification network architectures. We also show that scaling to even seven times as much training data with standard loss cannot beat the accuracy of CYBORG loss. As a side effect, we observed that the addition of explicit region annotation to the task of synthetic face detection increased human classification performance. This work opens a new area of research on how to incorporate human visual saliency into loss functions. All data, code and pre-trained models used in this work are offered with this paper.

The Value of AI Guidance in Human Examination of Synthetically-Generated Faces

“…Why the model came to this decision, on the other hand, is more complex. The interpretation in this work of why the AI made that decision relates to areas in the image We use pre-trained DenseNet-121 models to detect synthetic faces (Boyd et al 2021), and devise three separate experimental settings, as illustrated in Fig. 3.…”

“…CAMs are overlayed on the original image as a heatmap, where red and blue regions represent more and less salient regions, respectively. Because we could use 10 models in each training setting (Boyd et al 2021), we combine the generated CAM heatmaps into a GIF that smoothly transitions between all 10 unique CAMs over a 10s period.…”

“…In this "partnership", both humans and AI guide their counterparts to salient information, helping them to understand the problem and ultimately make better decisions. In (Boyd, Bowyer, and Czajka 2022;Boyd et al 2021), the first step of human-AI supervision (humans guiding the AI) was implemented for the task of synthetic iris and face detection. Human annotations of salient image regions were either used to augment the training data, or were incorporated into the loss function to guide the network's training process, resulting in better generalization against unseen image types.…”

Deep Learning for Iris Recognition: A Survey

“…Human-aided DL training is another promising avenue. Indeed, humans and machines cooperating in vision tasks is not new, and this strategy is finding its way into DL as well [14,17]. For example, Boyd et al [14] analyzed the utility of human judgement about salient regions of images to improve generalization of DL models.…”

“…Asked about regions that humans deem important for their decision about an image, the work proposed to transform the training data to incorporate such opinions, demonstrating an improvement in accuracy and generalization in leave-one-attack-type-out scenarios. In a similar work, Boyd et al [17] incorporated annotated saliency maps into the loss function to penalize large differences with human judgement.…”