Update Semantic Segmentation app

[wxm2018]

newly add a SelfSemanticSegmentationTranslator class to get the original photo pixels, modify the way of model loading in SemanticModel.java, and add a SwitchCompat in SemanticActivity.java

[frankfliu]

@wxm2018
Thanks a lot for your contribution, I think the right way to improve the SemanticSegmentation is to improve the SemanticSegmentationTranslator directly. Would you mind apply your change directly to: https://github.com/deepjavalibrary/djl/blob/master/api/src/main/java/ai/djl/modality/cv/translator/SemanticSegmentationTranslator.java

[wxm2018]

@frankfliu Ok, I will resubmit a new PR. And I discussed this with xyang. Her suggestion was to keep the original Translator. Can I add a new custom xxxTranslator.java file in the ai.djl.modality.cv.translator package?

[lanking520]

Can current implementation crop the drawing area?

[xyang16]

Can current implementation crop the drawing area?

You mean only remove the background?

[lanking520]

Like remove background and narrow to the area we need

[xyang16]

Like remove background and narrow to the area we need

We need to keep the whole background (to replace to other color or image)

[xyang16]

Update app. See the screenshot below.