OMNIA Faster R-CNN: Detection in the wild through dataset merging and soft distillation

Abstract

Object detectors tend to perform poorly in new or open domains, and require exhaustive yet costly annotations from fully labeled datasets. We aim at benefiting from several datasets with different categories but without additional labelling, not only to increase the number of categories detected, but also to take advantage from transfer learning and to enhance domain independence. Our dataset merging procedure starts with training several initial Faster R-CNN on the different datasets while considering the complementary datasets' images for domain adaptation. Similarly to self-training methods, the predictions of these initial detectors mitigate the missing annotations on the complementary datasets. The final OMNIA Faster R-CNN is trained with all categories on the union of the datasets enriched by predictions. The joint training handles unsafe targets with a new classification loss called SoftSig in a softly supervised way. Experimental results show that in the case of fashion detection for images in the wild, merging Modanet with COCO increases the final performance from 45.5% to 57.4% in mAP. Applying our soft distillation to the task of detection with domain shift between GTA and Cityscapes enables to beat the state-of-the-art by 5.3 points. Our methodology could unlock object detection for real-world applications without immense datasets.

arXiv comment9 pages, 5 figures, 4 tables

References in corpus (18)

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• Distilling the Knowledge in a Neural Network
• R-FCN: Object Detection via Region-based Fully Convolutional Networks
• Learning Transferable Features with Deep Adaptation Networks
• Deep Visual Domain Adaptation: A Survey
• Focal Loss for Dense Object Detection
• Training Deep Neural Networks on Noisy Labels with Bootstrapping
• Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes
• Unifying distillation and privileged information
• An Implementation of Faster RCNN with Study for Region Sampling
• Zero-Shot Detection
• Soft Sampling for Robust Object Detection
• Data Distillation: Towards Omni-Supervised Learning
• Novelty Detection with GAN
• ModaNet: A Large-Scale Street Fashion Dataset with Polygon Annotations
• Open Set Domain Adaptation by Backpropagation
• A Two-Stream Mutual Attention Network for Semi-supervised Biomedical Segmentation with Noisy Labels
• Pixel and Feature Level Based Domain Adaption for Object Detection in Autonomous Driving

Cited by in corpus (5)

• Domain Adaptation for Object Detection via Style Consistency
• Active learning using weakly supervised signals for quality inspection
• Self-supervised Robust Object Detectors from Partially Labelled Datasets
• Dynamic Supervisor for Cross-dataset Object Detection
• Model Composition: Can Multiple Neural Networks Be Combined into a Single Network Using Only Unlabeled Data?