TESTING_ON_GENERAL_DATASET.md

Testing on a General Dataset

Before we begin, make sure you have followed the steps here to setup your dataset.

This README aims to document the additional steps required to test/evaluate a general dataset. If you would like to evaluate it instead, see the document for training a general dataset.

Note that there are 2 types of "testing" that you can do:

1. Evaluation: you have ground truth keypoints data, and what to evaluate against this ground truth
2. Demo: you do not have ground truth keypoints, and want to use this algorithm to generate these keypoints

Evaluation

Simply run

python3 train.py \
  --eval --eval_dataset val \
  --config experiments/path/to/config_file.yaml \
  --logdir ./logs

Argument --eval_dataset can be val or train. Results can be seen in logs directory or in the tensorboard.

Demo

Simply run

python3 demo.py \
  --config experiments/path/to/config_file.yaml \
  --logdir ./logs

Argument --eval_dataset can be val or train. Results can be seen in logs directory or in the tensorboard.

Visualising Results

To visualise your results, follow the instructions in the README. You can choose to run with or [without]((README.md#visualising-results-without-tensorboard) tensorboard.