[论文解读] An Efficient Domain-Incremental Learning Approach to Drive in All Weather Conditions

M. Jehanzeb Mirza, Marc Masana+2 more|arXiv (Cornell University)|Apr 19, 2022

Domain Adaptation and Few-Shot Learning被引用 9

一句话总结

本文提出DISC，一种用于自动驾驶的领域增量学习方法，通过仅存储每种天气条件下的领域特定一阶和二阶统计量（均值和方差），实现零遗忘。通过在推理时动态替换冻结检测模型中的这些统计量，DISC能够在无需微调或存储模型权重的情况下，实时适应雨、雾、雪等新天气条件，以极低的内存和计算成本在所有天气场景下实现鲁棒的物体检测。

ABSTRACT

Although deep neural networks enable impressive visual perception performance for autonomous driving, their robustness to varying weather conditions still requires attention. When adapting these models for changed environments, such as different weather conditions, they are prone to forgetting previously learned information. This catastrophic forgetting is typically addressed via incremental learning approaches which usually re-train the model by either keeping a memory bank of training samples or keeping a copy of the entire model or model parameters for each scenario. While these approaches show impressive results, they can be prone to scalability issues and their applicability for autonomous driving in all weather conditions has not been shown. In this paper we propose DISC -- Domain Incremental through Statistical Correction -- a simple online zero-forgetting approach which can incrementally learn new tasks (i.e weather conditions) without requiring re-training or expensive memory banks. The only information we store for each task are the statistical parameters as we categorize each domain by the change in first and second order statistics. Thus, as each task arrives, we simply 'plug and play' the statistical vectors for the corresponding task into the model and it immediately starts to perform well on that task. We show the efficacy of our approach by testing it for object detection in a challenging domain-incremental autonomous driving scenario where we encounter different adverse weather conditions, such as heavy rain, fog, and snow.

研究动机与目标

• 解决深度学习模型在自动驾驶中适应新天气条件时出现的灾难性遗忘问题。
• 在不微调或存储大容量内存库的情况下，实现在多样化和变化的天气条件下的鲁棒物体检测。
• 开发一种可扩展、高效的增量学习方法，在保持对先前看到的天气条件性能的同时，适应新条件。
• 证明一种‘即插即用’的统计校正机制在真实世界、在线领域自适应中的可行性。

提出的方法

• 该方法通过检测模型特征图的一阶和二阶统计量（均值μ和方差σ²）的偏移来对每种天气条件进行分类。
• 在推理时，模型使用冻结的主干网络，仅将批量归一化统计量替换为对应当前天气条件的任务特定值。
• 该方法无需微调或存储模型参数；仅需为每种天气条件存储领域特定的统计量（μ_ID, σ²_ID）。
• 系统根据当前领域在推理时动态‘插入’正确的统计量，实现即时适应。
• 该方法以在线、增量方式运行，可在不降低先前条件性能的前提下，顺序适应多种天气变化。
• 统计校正仅在推理时应用，使该方法计算开销轻量，适合实时部署。

Figure 1 : Robust object detection for autonomous driving across varying weather conditions with DISC. We only require domain-specific statistics ( i.e . $\mu_{\text{ID}},\sigma_{\text{ID}}^{2}$ per weather condition ID) which are inserted into a frozen detection model to achieve zero-forgetting , i

实验结果

研究问题

• RQ1 领域增量学习方法能否在自动驾驶中适应新天气条件时实现零遗忘？
• RQ2 能否在不微调或存储模型参数的情况下，维持在多次天气变化下的物体检测性能？
• RQ3 对批量归一化统计量进行统计校正，在多样化天气条件下实现鲁棒推理的有效性如何？
• RQ4 该方法能否在重雨、浓雾和大雪等挑战性天气条件下实现泛化？

主要发现

• DISC实现了零遗忘，在适应新条件（如雾或雪）时，仍能保持对先前看到的天气条件（如晴天、雨天）的检测性能。
• 该方法表现出强迁移能力：雨天和雪天的统计量相比源模型基线，能提升其他条件下的性能。
• 仅在雾天数据上表现最佳的雾天特定统计量，仍优于晴天统计量在雾天的表现，表明领域特定适应是有效的。
• 与竞争的在线方法相比，DISC的性能下降显著更低，尤其在涉及大雪和浓雾的挑战性序列中。
• 跨任务评估表明，DISC的泛化能力优于独立模型，后者在不同天气条件间表现出差的迁移能力。
• 在类别级分析中，骑行人是浓雾中检测最困难的物体类别，凸显了未来研究的关键挑战。

Figure 2 : Domain-incremental learning setting used in this paper. Different tasks correspond to different weather conditions which are learned in the training phase. During inference, we evaluate the performance of the current task along with all the previously seen tasks.