Domain penalisation for improved Out-of-Distribution Generalisation
Shuvam Jena, Sushmetha Sumathi Rajendran, Karthik Seemakurthy, Sasithradevi A, Vijayalakshmi M, Prakash Poornachari
TL;DR
This work addresses domain generalisation for object detection across multiple source domains using GWHD 2021 as a benchmark. It introduces Domain Penalisation (DP), a dynamic, per-domain loss weighting scheme with $L_B = \sum_i w_i L_{ij}$, $L_{D_i} = \frac{1}{M_i} \sum_j L_{ij}$ and $w_i = \frac{\exp(L_{D_i})}{\sum_k \exp(L_{D_k})}$, and initial weights $w_i=1$. The approach is detector-agnostic and evaluated on FasterRCNN and FCOS, where DP outperforms ERM and GroupDRO in out-of-distribution accuracy on GWHD 2021. DP also reduces the ID–OOD generalisation gap across many domains, indicating more stable transfer to unseen environments, and the method can be extended to other WiLDS datasets and, potentially, to image classification tasks.
Abstract
In the field of object detection, domain generalisation (DG) aims to ensure robust performance across diverse and unseen target domains by learning the robust domain-invariant features corresponding to the objects of interest across multiple source domains. While there are many approaches established for performing DG for the task of classification, there has been a very little focus on object detection. In this paper, we propose a domain penalisation (DP) framework for the task of object detection, where the data is assumed to be sampled from multiple source domains and tested on completely unseen test domains. We assign penalisation weights to each domain, with the values updated based on the detection networks performance on the respective source domains. By prioritising the domains that needs more attention, our approach effectively balances the training process. We evaluate our solution on the GWHD 2021 dataset, a component of the WiLDS benchmark and we compare against ERM and GroupDRO as these are primarily loss function based. Our extensive experimental results reveals that the proposed approach improves the accuracy by 0.3 percent and 0.5 percent on validation and test out-of-distribution (OOD) sets, respectively for FasterRCNN. We also compare the performance of our approach on FCOS detector and show that our approach improves the baseline OOD performance over the existing approaches by 1.3 percent and 1.4 percent on validation and test sets, respectively. This study underscores the potential of performance based domain penalisation in enhancing the generalisation ability of object detection models across diverse environments.