Introspective False Negative Prediction for Black-Box Object Detectors in Autonomous Driving

Abstract

Object detection plays a critical role in autonomous driving, but current state-of-the-art object detectors will inevitably fail in many driving scenes, which is unacceptable for safety-critical automated vehicles. Given the complexity of the real traffic scenarios, it is impractical to guarantee zero detection failure; thus, online failure prediction is of crucial importance to mitigate the risk of traffic accidents. Of all the failure cases, False Negative (FN) objects are most likely to cause catastrophic consequences, but little attention has been paid to the online FN prediction. In this paper, we propose a general introspection framework that can make online prediction of FN objects for black-box object detectors. In contrast to existing methods which rely on empirical assumptions or handcrafted features, we facilitate the FN feature extraction by an introspective FN predictor we designed in this framework. For this purpose, we extend the original concept of introspection to object-wise FN predictions, and propose a multi-branch cooperation mechanism to address the distinct foreground-background imbalance problem of FN objects. The effectiveness of the proposed framework is verified through extensive experiments and analysis, and the results show that our method successfully predicts the FN objects with 81.95% precision for 88.10% recall on the challenging KITTI Benchmark, and effectively improves object detection performance by taking FN predictions into consideration.

Run-time Introspection of 2D Object Detection in Automated Driving Systems Using Learning Representations

“…Extensions of this approach involve monitoring sequences of frames or employing cascaded networks to enhance detection of performance drops [26]. Additionally, Yang et al propose a method to predict object-level false negatives using an introspection model independent of the underlying object detector [9].…”

“…For instance, in [8], the intermediate states of object detection networks, also known as activation maps, are processed and used to detect erroneous predictions based on the model's performance on a selected metric, such as the mean average precision (mAP). Alternatively, in [9], the input image is directly utilised and fed into an object detector-like architecture to predict missed objects. Also, in [10], confidence values are combined with image features, such as the entropy of the colour histogram to discover erroneous cases.…”

What's in the Black Box? The False Negative Mechanisms Inside Object Detectors

“…In the last few years, a number of works have emerged introducing techniques that identify when an object detector fails to detect an object [21]- [23]. These works all highlighted the danger of false negatives in object detectors aboard autonomous systems, arguing potentially catastrophic consequences in applications such as autonomous driving [21]- [23].…”

“…In the last few years, a number of works have emerged introducing techniques that identify when an object detector fails to detect an object [21]- [23]. These works all highlighted the danger of false negatives in object detectors aboard autonomous systems, arguing potentially catastrophic consequences in applications such as autonomous driving [21]- [23]. Without any existing research informing why object detectors produce false negatives, these works instead focused on exploiting other signals for detecting false negatives -temporal or stereo camera detection inconsistencies [22], learning underlying biases in the objects that produce false negatives [23], and relying on hand-crafted indicators of false negatives in a detector's feature maps [21].…”

Driving Behavior Classification and Sharing System Using CNN-LSTM Approaches and V2X Communication

“…Despite advances in autonomous driving technology, traffic accidents remain a problem to be solved in the transportation system [13,14]. According to the World Health Organization, statistics are reported that ~1.3 million people die in traffic accidents each year [15].…”