A simple baseline for 3d human pose estimation in tensorflow. Presented at ICCV 17.
Intrinsic3D - High-Quality 3D Reconstruction by Joint Appearance and Geometry Optimization with Spatially-Varying Lighting (ICCV 2017)
翻译 - Intrinsic3D-通过空间变化照明的联合外观和几何优化实现高质量3D重构（ICCV 2017）
RMPE: Regional Multi-person Pose Estimation, forked from Caffe. Research purpose only.
Soft Proposal Networks for Weakly Supervised Object Localization, in ICCV 2017
#计算机科学#Code for "Unsupervised Adaptation for Deep Stereo" - ICCV17
#计算机科学#Chainer Implementation of Mask R-CNN. (Training code to reproduce the original result is available.)
this repo attemps to reproduce Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks(CycleGAN) use gluon reimplementation
this repo attemps to reproduce DSOD: Learning Deeply Supervised Object Detectors from Scratch use gluon reimplementation
#计算机科学#This is an un-official re-implementation of the work Sketching with Style presented at ICCV 2017 by Prof. John Collomosse's Group and Adobe.
近期要读的2017年顶会的几篇行为识别论文