ChainerCV: a Library for Deep Learning in Computer Vision

• DOI: 10.48550/arxiv.1708.08169
• Publication Date: 2017-01-01
• Authors: Yusuke Niitani; Toru Ogawa; Shunta Saito; Masaki Saito
• Keywords: Implementation; Simplicity

Despite significant progress of deep learning in the field computer vision, there has not been a software library that covers these methods unifying manner. We introduce ChainerCV, is intended to fill this gap. ChainerCV supports numerous neural network models as well components needed conduct research vision. These implementations emphasize simplicity, flexibility and good engineering practices. The designed perform on par with results reported published papers its tools can be used baseline for future Our implementation includes sophisticated like Faster R-CNN SSD, tasks such object detection semantic segmentation.