MultiNet

Why we included this project

MultiNet is a compact example of how to run several driving-related vision tasks in one network. It shares a single VGG encoder across independent decoders, so road segmentation, car detection, and street classification happen together instead of as separate models, which keeps inference fast enough for real-time use. The code pulls in KittiSeg, KittiBox, and KittiClass as submodules, and the demo script can produce a prediction on a single image after a short setup. The accompanying paper explains the architecture, and the TensorVision-based experiment management shows a clean way to track training runs and checkpoints. Just keep in mind the project targets Python 2.7 and TensorFlow 1.0, so treat it as a learning and prototyping reference rather than something to drop into a modern stack unchanged.