Tiny-SSD

An implementation of SSD

Usage

1. Download voc2007 or voc2012 or voc2077(My small toy face detection dataset, you can use any dataset which follows the VOC structure)
2. Put them in the right directory
3. Run voc_annotation to generate divided dataset(in VOCdevkit\VOC20..\ImageSets\Main) and dataset labels(in code's main directory)
4. Run train.py to train a simple model(will not consume too many time). It will save .pt weights in model_date/ every epoch automatically
5. inference or evaluation(config info can be modified in model/tiny_ssd.py):
   1. run inference.py to get object detection result from a picture or your pc camera
   2. run evaluation.py to evaluate your test set

Result(My small toy face detection results)

Reference

1. Dive into Deep Learning
2. An implementation of YOLOv3