gin commit from weian-pro-28

New files: 4
Modified files: 1

lightningpose/README.md

@@ -6,5 +6,23 @@ https://doi.org/10.1101/2023.04.28.538703) pose estimation algorithm on the exam
 The example data is located here: https://github.com/danbider/lightning-pose/tree/main/data/mirror-mouse-example
 The output was generated from this [notebook](https://colab.research.google.com/github/danbider/lightning-pose/blob/main/scripts/litpose_training_demo.ipynb)
 
+### Example directory structure 
+
+```
+/outputs/YYYY-MM-DD/HH-MM-SS/
+  └── video_preds/
+      ├── labeled_videos/
+      ├── test_vid.csv
+      ├── test_vid_pca_multiview_error.csv
+      ├── test_vid_pca_singleview_error.csv
+      └── test_vid_temporal_norm.csv
+```
+
+* `video_preds/`: contains the predictions and metrics from videos.
+* `video_preds/labeled_videos/`: the labeled mp4 videos
+* `video_preds/test_vid.mp4`: the original video
+* `video_preds/test_vid.csv`: the predictions on labeled data (x, y coordinates and confidence)
+* `video_preds/test_vid_pca_multiview_error.csv`: the pca multiview reprojection error between predictions and labeled keypoints
+* `video_preds/test_vid_pca_singleview_error.csv`: pca singleview reprojection error between predictions and labeled keypoints
+* `video_preds/test_vid_temporal_norm.csv`: the temporal norm between predictions and labeled key points
 
-The format consists of the original video in mp4 format that was used for predictions, the CSV file that contains the model output (x, y coordinates and confidence), the model metrics in CSV files (e.g. multi-view PCA, temporal norm, pose PCA) and optionally the labeled video in mp4 format.