Explainable Multi-Camera 3D Object Detection with Transformer-Based Saliency Maps
This work addresses the need for transparency in safety-critical AI applications like autonomous driving, though it is incremental as it builds on existing transformer-based detection methods.
The paper tackles the problem of explainability in Vision Transformers for 3D object detection in autonomous driving by proposing a novel method to generate saliency maps based on raw attention, showing it outperforms other methods on the nuScenes dataset in visual quality and quantitative metrics.
Vision Transformers (ViTs) have achieved state-of-the-art results on various computer vision tasks, including 3D object detection. However, their end-to-end implementation also makes ViTs less explainable, which can be a challenge for deploying them in safety-critical applications, such as autonomous driving, where it is important for authorities, developers, and users to understand the model's reasoning behind its predictions. In this paper, we propose a novel method for generating saliency maps for a DetR-like ViT with multiple camera inputs used for 3D object detection. Our method is based on the raw attention and is more efficient than gradient-based methods. We evaluate the proposed method on the nuScenes dataset using extensive perturbation tests and show that it outperforms other explainability methods in terms of visual quality and quantitative metrics. We also demonstrate the importance of aggregating attention across different layers of the transformer. Our work contributes to the development of explainable AI for ViTs, which can help increase trust in AI applications by establishing more transparency regarding the inner workings of AI models.