MULTICLOD: Multiclass object detection

Computer vision for smart cars and safer roads


Ladder-style DenseNet for semantic segmentation

  • TensorFlow implementation of a semantic segmentation approach which leverages ladder-style lateral connections to blend spatial resolution from the early layers with semantic expressiveness from the deep layers of a modified DenseNet architecture
  • Coming soon! Work in progress

Scale invariant semantic segmentation

  • this is a Torch implementation of an approach which recovers a scale invariant convolutional representation by leveraging stereo reconstruction to decouple appearance from scale


  • This our semantic annotator, a tool for recovering semantic segmentation groundtruth
  • we used the tool to annotate 300 hand-picked images from KITTI (the iamges are available here).