ObjectDetection.java using other backbone than resnet50

[TjorvenDM]

I'm fairly new to the djl library and I am looking at the example of ObjectDetection.java but have a hard time understanding why i can only use resnet50 as backbone although there are others specified (see image). I'm probably missing something here, if someone could help me out I'd appreciate it a lot!

[frankfliu]

I'm able to load vgg16 ssd model:

        Criteria<Image, DetectedObjects> criteria =
                Criteria.builder()
                        .optApplication(Application.CV.OBJECT_DETECTION)
                        .setTypes(Image.class, DetectedObjects.class)
                        .optFilter("backbone", "vgg16")
                        .optFilter("dataset", "voc")
                        .optEngine("MXNet")
                        .optProgress(new ProgressBar())
                        .build();

[frankfliu]

For different backbone, the flavor also different, to list the models, you can run the following command:

./gradlew listmodels

[INFO ] - CV.OBJECT_DETECTION ai.djl.mxnet/ssd/0.0.1/ssd_512_resnet50_v1_voc {"size":"512","backbone":"resnet50","flavor":"v1","dataset":"voc"}
[INFO ] - CV.OBJECT_DETECTION ai.djl.mxnet/ssd/0.0.1/ssd_512_vgg16_atrous_coco {"size":"512","backbone":"vgg16","flavor":"atrous","dataset":"coco"}
[INFO ] - CV.OBJECT_DETECTION ai.djl.mxnet/ssd/0.0.1/ssd_512_mobilenet1.0_voc {"size":"512","backbone":"mobilenet1.0","dataset":"voc"}
[INFO ] - CV.OBJECT_DETECTION ai.djl.mxnet/ssd/0.0.1/ssd_300_vgg16_atrous_voc {"size":"300","backbone":"vgg16","flavor":"atrous","dataset":"voc"}