VisualBackProp: Efficient Visualization of CNNs for Autonomous Driving

Abstract

This paper proposes a new method, that we call VisualBackProp, for visualizing which sets of pixels of the input image contribute most to the predictions made by the convolutional neural network (CNN). The method heavily hinges on exploring the intuition that the feature maps contain less and less irrelevant information to the prediction decision when moving deeper into the network. The technique we propose was developed as a debugging tool for CNN-based systems for steering self-driving cars and is therefore required to run in real-time, i.e. it was designed to require less computations than a forward propagation. This makes the presented visualization method a valuable debugging tool which can be easily used during both training and inference. We furthermore justify our approach with theoretical arguments and theoretically confirm that the proposed method identifies sets of input pixels, rather than individual pixels, that collaboratively contribute to the prediction. Our theoretical findings stand in agreement with the experimental results. The empirical evaluation shows the plausibility of the proposed approach on the road video data as well as in other applications and reveals that it compares favorably to the layer-wise relevance propagation approach, i.e. it obtains similar visualization results and simultaneously achieves order of magnitude speed-ups.

Grounding Human-To-Vehicle Advice for Self-Driving Vehicles

“…We visualize raw input images with salient objects marked by a green circle, e.g., a bus pulling off, which is mentioned by an advice input (1st row). The provided advice (1)(2)(3)(4)(5)(6) is provided at the bottom of the figure. We visualize attention heat maps from our trained model but with a synthetic token <none> (i.e., without advice, 2nd row).…”

“…The recent achievements [3,27] suggest that deep neural models can be applied to vehicle controls in an end-to-end manner by effectively learning latent representations from data. Explainability of these deep controllers has increasingly been explored via a visual attention mechanism [8], a deconvolution-style approach [2], and a natural language model [9]. Such explainable models will be an important element of human-vehicle interaction because they allow people and vehicles to understand and anticipate each other's actions, hence to cooperate effectively.…”

A survey of deep learning techniques for autonomous driving

“…An autonomy of was reached on a 20‐km drive from Holmdel to Atlantic Highlands, NJ. Through training, PilotNet learns how the steering commands are computed by a human driver (Bojarski et al, ). The focus is on determining which elements in the input traffic image have the most influence on the network's steering decision.…”

Towards Automated Melanoma Detection With Deep Learning: Data Purification and Augmentation

“…they use visual cues such as hairs and rulers as indicators of the lesion category. To demonstrate this problem we employ the visualization technique, called Vi-sualBackProp [9], that highlights the part of the image that the network focuses on when forming its prediction. Figure 2 shows the results obtained for the traditionally-trained deep model (without performing data purification or augmentation) on the raw data.…”