Abstract:The development of deep learning techniques has greatly benefited CNN-based object detectors, leading to unprecedented progress in recent years. However, the distribution variance between training and testing domains causes significant performance degradation. Labeling data for new scenarios is costly and time-consuming, so most existing domain adaptation methods perform feature alignment through adversarial training. While this can improve the accuracy of detectors in unlabeled target domains, the unconstrained domain alignment also negatively transfers the feature distribution, which compromises the recognition ability of the model. To address this problem, we propose the Knowledge Transfer Network (KTNet), which consists of object intrinsic knowledge mining and category relational knowledge constraint modules. Specifically, a binary classifier shared by the source and target domains is designed to extract common attribute knowledge of objects, which can align foreground and background features from different data domains adaptively. Then, we construct relational knowledge graphs to explicitly constrain the category correlations in the source, target, and cross-domain settings. These two modules guide the detector to learn object-related and domain-invariant representations, enabling the proposed KTNet to perform well in four commonly-used cross-domain scenarios. Furthermore, the ablation experiments show that our method is scalable to more complex backbone networks and different detection architectures.

Deformable Part-Based Model Transfer for Object Detection.

Assemble New Object Detector With Transferred Auxiliary Filters And Distinct Target Filters

Improved Deformable Part Model for Object Detection Based on Scale Invariant Feature Transform

Object Detection in VHR Image Using Transfer Learning with Deformable Convolution

An Empirical Study of Deformable Part Model with Fast Feature Pyramid

Object Detection with Discriminatively Trained Part Based Models

Active Deformable Part Models

DSN: A New Deformable Subnetwork for Object Detection

Balanced Mixture of Deformable Part Models with Automatic Part Configurations.

A DPM Based Approach to Joint Object Detection and Sub-category Recognition

Parallelized Deformable Part Models with Effective Hypothesis Pruning.

Domain Adaptation of Deformable Part-Based Models

Object Tracking with Embedded Deformable Parts in Dynamic Conditional Random Fields

Enhancing Adversarial Transferability in Object Detection with Bidirectional Feature Distortion.

PM-DETR: Domain Adaptive Prompt Memory for Object Detection with Transformers

Bayesian Transfer Learning for Object Detection in Optical Remote Sensing Images

Deformable Part Models are Convolutional Neural Networks

Region Feature Disentanglement for Domain Adaptive Object Detection

Deformable DETR: Deformable Transformers for End-to-End Object Detection

Domain adaptive object detection with model-agnostic knowledge transferring

DeepID-Net: Deformable Deep Convolutional Neural Networks for Object Detection