SALAD: Source-free Active Label-Agnostic Domain Adaptation for Classification, Segmentation and Detection

TL;DR

The paper tackles the challenge of source-free active domain adaptation under a limited target annotation budget and potential label-space shifts. It introduces SALAD, a framework that couples a Guided Attention Transfer Network (GATN) for feature-level distillation from a frozen source model with an active learning strategy (H_AL) that selects informative target samples to annotate. Across classification, segmentation, and document layout detection tasks, SALAD matches or exceeds performance of baselines that rely on substantial annotated source data, with notably strong gains in CityScapes segmentation and DSSE document layout tasks using very small target budgets. This work demonstrates SALAD's practical value in privacy-preserving settings and its broad applicability to multiple vision tasks without access to source data during adaptation.

Abstract

We present a novel method, SALAD, for the challenging vision task of adapting a pre-trained "source" domain network to a "target" domain, with a small budget for annotation in the "target" domain and a shift in the label space. Further, the task assumes that the source data is not available for adaptation, due to privacy concerns or otherwise. We postulate that such systems need to jointly optimize the dual task of (i) selecting fixed number of samples from the target domain for annotation and (ii) transfer of knowledge from the pre-trained network to the target domain. To do this, SALAD consists of a novel Guided Attention Transfer Network (GATN) and an active learning function, HAL. The GATN enables feature distillation from pre-trained network to the target network, complemented with the target samples mined by HAL using transfer-ability and uncertainty criteria. SALAD has three key benefits: (i) it is task-agnostic, and can be applied across various visual tasks such as classification, segmentation and detection; (ii) it can handle shifts in output label space from the pre-trained source network to the target domain; (iii) it does not require access to source data for adaptation. We conduct extensive experiments across 3 visual tasks, viz. digits classification (MNIST, SVHN, VISDA), synthetic (GTA5) to real (CityScapes) image segmentation, and document layout detection (PubLayNet to DSSE). We show that our source-free approach, SALAD, results in an improvement of 0.5%-31.3%(across datasets and tasks) over prior adaptation methods that assume access to large amounts of annotated source data for adaptation.

Figures (1)

• Figure 1: We present a generic method, SALAD, for the task of adapting from a pre-trained source network to a target domain with a small budget for acquiring annotation, under a possible shift in label space. SALAD consists of two complementary components: Guided Attention Transfer Network (GATN) and an active learning strategy $H_{AL}$.