Seeing in the Dark: A Teacher-Student Framework for Dark Video Action Recognition via Knowledge Distillation and Contrastive Learning

TL;DR

This paper introduces ActLumos, a teacher-student framework that leverages contrastive learning and knowledge distillation to improve dark video action recognition accuracy while maintaining single-stream inference efficiency.

Contribution

It proposes a novel teacher-student architecture with dynamic feature fusion and contrastive loss, achieving state-of-the-art results in dark video action recognition.

Findings

Student model achieves 96.92% Top-1 accuracy on ARID V1.0

Dynamic feature fusion outperforms static fusion methods

Two-view SSL surpasses spatial-only or temporal-only variants

Abstract

Action recognition in dark or low-light (under-exposed) videos is a challenging task due to visibility degradation, which can hinder critical spatiotemporal details. This paper proposes ActLumos, a teacher-student framework that attains single-stream inference while retaining multi-stream level accuracy. The teacher consumes dual stream inputs, which include original dark frames and retinex-enhanced frames, processed by weight-shared R(2+1)D-34 backbones and fused by a Dynamic Feature Fusion (DFF) module, which dynamically re-weights the two streams at each time step, emphasising the most informative temporal segments. The teacher is also included with a supervised contrastive loss (SupCon) that sharpens class margins. The student shares the R(2+1)D-34 backbone but uses only dark frames and no fusion at test time. The student is first pre-trained with self-supervision on dark clips of both datasets without their labels and then fine-tuned with knowledge distillation from the teacher, transferring the teacher's multi-stream knowledge into a single-stream model. Under single-stream inference, the distilled student attains state-of-the-art accuracy of 96.92% (Top-1) on ARID V1.0, 88.27% on ARID V1.5, and 48.96% on Dark48. Ablation studies further highlight the individual contributions of each component, i.e., DFF in the teacher outperforms single or static fusion, knowledge distillation (KD) transfers these gains to the single-stream student, and two-view spatio-temporal SSL surpasses spatial-only or temporal-only variants without increasing inference cost. The official website of this work is available at: https://github.com/HrishavBakulBarua/ActLumos