can train and predict text classification via BERT embeddings

adding factory to create bbox datasets

fix unit tests for having label in bbox output

fix bounding box to output labels

fix bbox resize speed by parallelization across threads vs processes

add resize and plot commands for bbox

run formatter, and add tsv and csv export for human keypoints

refactor a ton. have tool to parse coco bounding boxes to tsv or csv and unit tests

refactoring and cleaning up human pose annotation classes

reorganizing repo and adding unit tests

add ability to scale with different width and height

adding plot example

add sanity checks for number of keypoints on tsv loading

sanity check on input training data

skip empty lines in dataloader