Hello, glad you liked the article! I do see it being used in sports for applications such as remote/advanced training sessions, automatically finding penalties, posture correction and so on. I’ll consider augmenting the application section in the future or in a separate blog. Thanks for the suggestions!
Hello, glad you liked the article! I would suggest to go with an approach that is well documented and is reasonably fast, as most of these approaches have pretty good accuracy. With that being said, I believe OpenPose (or its optimized variants) would be a good place to start.
Glad you liked the article!
Regarding your question, consider the previous paragraph where it was stated “Now, say our model is 80% sure about pixels that are background, but only 30% sure about pixels that are the target class.” In the cross-entropy graph posted above that paragraph, when probability of a class was 0.8…
Hello! Glad you liked the article.
If a distribution has low entropy, it is very confident about an outcome (there is very little scope for randomness). If a distribution has high entropy, the outcome is very random (like an uniform distribution).