Deep CNN Ensemble with Data Augmentation for Object Detection

Citations Over Time

Abstract

We report on the methods used in our recent DeepEnsembleCoco submission to the PASCAL VOC 2012 challenge, which achieves state-of-the-art performance on the object detection task. Our method is a variant of the R-CNN model proposed Girshick:CVPR14 with two key improvements to training and evaluation. First, our method constructs an ensemble of deep CNN models with different architectures that are complementary to each other. Second, we augment the PASCAL VOC training set with images from the Microsoft COCO dataset to significantly enlarge the amount training data. Importantly, we select a subset of the Microsoft COCO images to be consistent with the PASCAL VOC task. Results on the PASCAL VOC evaluation server show that our proposed method outperform all previous methods on the PASCAL VOC 2012 detection task at time of submission.

Related Papers

• → Detecting Avocados to Zucchinis: What Have We Done, and Where Are We Going?(2013)88 cited
• → Object detection through region proposal based techniques(2021)13 cited
• → Object detection using improved YOLOv3-tiny based on pyramid pooling(2021)1 cited
• → Improving object detection via improving accuracy of object localization(2016)
• → Improved Regional Proposal Generation and Proposal Selection Method for Weakly Supervision Object detection(2023)