Code to detect keypoints has been updated with framework proposed by MIT to get a better performance. In addition, it has been
added and script to optimize hiperparameters by a bayesian optimization framework. To train it, it is needed a folder with images of cones labeled.
It think it's worth numpyfying calculate_distance and euclidean_distance_loss to gain those precious precious milliseconds for production environment 🤓
Code to detect keypoints has been updated with framework proposed by MIT to get a better performance. In addition, it has been added and script to optimize hiperparameters by a bayesian optimization framework. To train it, it is needed a folder with images of cones labeled.