Object detection for self driving. DS502-1702 capstone project track 1. We use SSD-Tensorflow as pre-trained model.

Team member: Jin Sun, Xintong Xia

For more details please look at our Presentation Slides

Installation:

virtualenv env
source env/bin/activate
pip install -r requirements.txt

Train: load pre-trained model, train (currently not implemented), and save trained model

./scripts/run.sh train

If you are running at first time, you have to running training commands before testing and serving, because it downloads pretrained model.

Test: load trained model, batch predict on test set

./scripts/run.sh test

Serve: run model on a video

./scripts/run.sh serve