SSD Improvements

This repository is forked from the project SSD. I improve SSD from the following aspects:

1. add Resnet backbone (Resnet50 and Resnet101)
2. add attention mechanism (Squeeze-and-Excitation Module , Convolutional Block Attention Module)
3. add feature fusion module

Train

Setting Up Datasets

Pascal VOC

For Pascal VOC dataset, make the folder structure like this:

VOC_ROOT
|__ VOC2007
    |_ JPEGImages
    |_ Annotations
    |_ ImageSets
    |_ SegmentationClass
|__ VOC2012
    |_ JPEGImages
    |_ Annotations
    |_ ImageSets
    |_ SegmentationClass
|__ ...

Where VOC_ROOT default is datasets folder in current project, you can create symlinks to datasets or export VOC_ROOT="/path/to/voc_root".

COCO

For COCO dataset, make the folder structure like this:

COCO_ROOT
|__ annotations
    |_ instances_valminusminival2014.json
    |_ instances_minival2014.json
    |_ instances_train2014.json
    |_ instances_val2014.json
    |_ ...
|__ train2014
    |_ <im-1-name>.jpg
    |_ ...
    |_ <im-N-name>.jpg
|__ val2014
    |_ <im-1-name>.jpg
    |_ ...
    |_ <im-N-name>.jpg
|__ ...

Where COCO_ROOT default is datasets folder in current project, you can create symlinks to datasets or export COCO_ROOT="/path/to/coco_root".

Train sample 1

Configs:
    Backbone: Resnet50
    Input size: 300
    SE: False
    CBAM: False
    FUSION: False
Run:
    python train.py --config-file configs/resnet50_ssd300_voc0712.yaml

Train sample 2

Configs:
    Backbone: Resnet50
    Input size: 300
    SE: False
    CBAM: False
    FUSION: True
Run:
    python train.py --config-file configs/resnet50_ssd300_voc0712_feature_fusion.yaml

You can do custom training by modifying the .yaml files.