Scene Labeling

Papers

Learning hierarchical features for scene labeling

• intro: “Their approach comprised of densely computing multi-scale CNN features for each pixel and aggregating them over image regions upon which they are classified. However, their methodstill required the post-processing step of generating over-segmented regions, like superpixels, for obtaining the final segmentation result. Additionally, the CNNs used for multi-scale feature learning were not very deep with only three convolution layers.”
• paper: http://yann.lecun.com/exdb/publis/pdf/farabet-pami-13.pdf

Indoor Semantic Segmentation using depth information

• arxiv: http://arxiv.org/abs/1301.3572

Multi-modal unsupervised feature learning for rgb-d scene labeling

• paper: http://www3.ntu.edu.sg/home/wanggang/WangECCV2014.pdf

Using neon for Scene Recognition: Mini-Places2

• intro: This is an implementation of the deep residual network used for Mini-Places2 as described in He et. al., “Deep Residual Learning for Image Recognition”.
• blog: http://www.nervanasys.com/using-neon-for-scene-recognition-mini-places2/
• github: https://github.com/hunterlang/mpmz

Attend, Infer, Repeat: Fast Scene Understanding with Generative Models

• arxiv: http://arxiv.org/abs/1603.08575

Challenges

Large-scale Scene Understanding Challenge

• homepage: http://lsun.cs.princeton.edu/