This is a ship detection pipeline for the Ships in Satellite Imagery Dataset from Kaggle. It uses Sliding Windows + Non maximal supression for the object detection with a HoG + SVM approach for classification.
A HoG + SVM classifier is trained using the information contained in the dataset compressed file and stored in pickle format.
python ship_clf.py
The ship detector slides a window across an image pyramid and classifies each window as ship or not ship returning a set of bounding boxes that are filtered afterwards using a non maximal supression scheme.
python ship_detector.py
