Just make sure your $CAFFE_PATH environment variable points to your caffe installation, e.g.:
export CAFFE_PATH=~/caffe/
Edit train_ssd.py to include:
dataset_root = Path to your VOC annotated dataset.
labels = Array with the labels of the classes you want to train.
Pre-trained weights are automatically downloaded from https://github.com/chuanqi305/MobileNet-SSD.
To start training just run:
python train_ssd.py
The generated weights were validated using a Raspberry Pi and an Intel NCS2, using code from this awesome repository: https://github.com/PINTO0309/MobileNet-SSD-RealSense.
Bash scripts were retrieved from https://github.com/weiliu89/caffe and https://github.com/chuanqi305/MobileNet-SSD.
More customization coming soon...