lightning_pose.utils

lightning_pose.utils.cropzoom Module

`generate_cropped_labeled_frames`(...[, ...])
`generate_cropped_video`(video_path, ...)

`check_lists_equal`(list_1, list_2)
`remove_string_w_substring_from_list`(strings, ...)
`check_dataset`(dataset)
`get_image_tags`(pred_df)

`FiftyOneImagePlotter`(cfg[, ...])
`dfConverter`(df, keypoint_names)

Path handling functions.

`ckpt_path_from_base_path`(base_path, model_name)	Given a path to a hydra output with trained model, extract the model .ckpt file.
`check_if_semi_supervised`([losses_to_use])	Use config file to determine if model is semi-supervised.
`get_keypoint_names`([cfg, csv_file, header_rows])
`return_absolute_path`(possibly_relative_path)	Return absolute path from possibly relative path.
`return_absolute_data_paths`(data_cfg[, ...])	Generate absolute path for our example toy data.
`get_videos_in_dir`(video_dir[, view_names, ...])	Gather videos to process from a single directory.
`check_video_paths`(video_paths[, view_names])
`get_context_img_paths`(center_img_path)	Given the path to a center image frame, return paths of 5 context frames (n-2, n-1, n, n+1, n+2).

PCA class to assist with computing PCA losses.

Reformat multiview data so each observation is a single body part across views.

`KeypointPCA`(loss_type, data_module[, ...])	Class to collect data from a dataloader and compute PCA params.
`ComponentChooser`(fitted_pca_object, ...)	Determine the number of PCA components to keep.

Functions for predicting keypoints on labeled datasets and unlabeled videos.

`predict_dataset`(cfg, data_module, preds_file)	Save predicted keypoints for a labeled dataset.
`predict_single_video`(cfg_file, video_file, ...)	Make predictions for a single video, loading frame sequences using DALI.
`make_dlc_pandas_index`(cfg, keypoint_names)
`get_model_class`(map_type, semi_supervised)	[summary]
`load_model_from_checkpoint`(cfg, ckpt_file[, ...])	Load Lightning Pose model from checkpoint file.
`create_labeled_video`(clip, xs_arr, ys_arr[, ...])	Helper function for creating annotated videos. Args clip xs_arr: shape T x n_joints ys_arr: shape T x n_joints mask_array: shape T x n_joints; timepoints/joints with a False entry will not be plotted dotsize: size of marker dot on labeled video colormap: matplotlib color map for markers fps: None to default to fps of original video output_video_path: video file name start_time: time (in seconds) of video start.
`export_predictions_and_labeled_video`(...[, ...])	Export predictions csv and a labeled video for a single video file.

PredictionHandler(cfg[, data_module, video_file])

Convert batches of model outputs into a prediction dataframe.

Helper functions to build pipeline components from config dictionary.

`get_imgaug_transform`(cfg)	Create simple data transform pipeline that augments images.
`get_dataset`(cfg, data_dir, imgaug_transform)	Create a dataset that contains labeled data.
`get_data_module`(cfg, dataset[, video_dir])	Create a data module that splits a dataset into train/val/test iterators.
`get_loss_factories`(cfg, data_module)	Create loss factory that orchestrates different losses during training.
`get_model`(cfg, data_module, loss_factories)	Create model: regression or heatmap based, supervised or semi-supervised.
`get_callbacks`(cfg[, early_stopping, ...])
`calculate_train_batches`(cfg[, dataset])	For semi-supervised models, this tells us how many batches to take from each dataloader (labeled and unlabeled) during a given epoch.
`compute_metrics`(cfg, preds_file[, data_module])	Compute various metrics on predictions csv file, potentially for multiple views.