Path to ndsampler demo

[banesullivan]

The following is a roadmap to putting together a demo where ndsampler uses ResonantGeoData to provide subsampled images.

For now, I will focus on a minimally working solution such that the core functionality of ndsampler can be used with RGD.

I will continually update this post as I work on these features

---

Roadmap:

• [ ] Create a Python API for hitting our endpoints.
  • Likely needed new endpoints:
  • [x] List all images associated with a KWCOCOArhcive
  • [x] List all annotations for a given image
  • [ ] Get annotation and segmentation
• [ ] Subclass BaseSampler -> new class RGDSampler
  • On init, use a query to define a list of images - fetch their annotations too. For simplicity at first, it might make sense to start with a KWCOCOArchive ID.
• [ ] implement load_sample(): See load_sample in CocoSampler docstring for input/output. Can ignore the annotations output for now.
• [ ] Implement the classes attribute of RGDSampler to return KWCOCO category tree
  • Likely requires us to implement regions
• [ ] Might also need to implement get_image, get_annotation, and get_category?
• [ ] Reproduce the example in the ndsampler readme
• [ ] Reproduce more complex examples
  • bioharn clf and detect
  • Netharn classification, object detection, and segmentaion

Other related tasks:

• [ ] add endpoints for uploading a KWCOCO dataset to RGD
• [ ] Add independent endpoints for uploading individual images and for adding annotations to images