MatConvNet-DeconvNet: A MatConvNet fork focused on deconvolutional neural networks. Tested for semantic segmentation, change detection, depth estimation, optical flow and more. It includes state-of-the-ar Adam optimizer.
Chainer-Deconv: A Chainer fork focused on deconvolutional neural networks. It includes code for training and testing semantic segmentation methods, including FCN and a demo of semantic segmentation in real-time with a webcam
SYNTHIA: A synthetic dataset for driving scenarios in a large world that includes features such as: dynamic objects (cars, pedestrians, cyclists), multiple seasons, dynamic lighting, pixel-wise groundtruth for semantic segmentation, depth, pose-estimation and more! Coming soon!