Model Directory Structure

/models/model_dir/
├── config.yaml
|
# Singleview
├── predictions.csv
# Multiview
├── predictions_view0.csv
├── predictions_view1.csv
|
├── tb_logs/ ...
├── exports_onnx/
│   └── epoch=214-step=12685-best_fp16.onnx
├── exports_trt/
│   └── epoch=214-step=12685-best_fp16/
│       └── trt_metadata.json
├── video_preds/
│   ├── session0_view0.mp4/
│   |   └── predictions.csv
│   └── session0_view1.mp4/
└── image_preds/
    └── CollectedData.csv/
         └── predictions.csv

Detailed descriptions

  • tb_logs/: model weights

  • exports_onnx/: ONNX exports produced by litpose export or Model.export("onnx", ...), one file per exported checkpoint and precision, named <checkpoint_stem>_<onnx_precision>.onnx. Read back by litpose predict --runtime onnx or Model.from_dir(..., runtime="onnx"). See Increasing Inference Speed.

  • exports_trt/: TensorRT engine caches produced by litpose export --runtime tensorrt or Model.export("tensorrt", ...), one directory per exported checkpoint and precision, named <checkpoint_stem>_<onnx_precision>/, built from the matching exports_onnx/ file. Read back by litpose predict --runtime tensorrt or Model.from_dir(..., runtime="tensorrt"). See Increasing Inference Speed.

  • video_preds/: predictions and metrics from videos. The config field eval.test_videos_directory points to a directory of videos; if eval.predict_vids_after_training is set to true, all videos in the indicated direcotry will be run through the model upon training completion and results stored here.

  • video_preds/labeled_videos/: labeled mp4s. The config field eval.test_videos_directory points to a directory of videos; if eval.save_vids_after_training is set to true, all videos in the indicated direcotry will be run through the model upon training completion and results stored here.

  • predictions.csv: predictions on labeled data. The right-most column records the train/val/test split that each example belongs to.

  • predictions_pixel_error.csv: Euclidean distance between the predictions in predictions.csv and the labeled keypoints (in <YOUR_LABELED_FRAMES>.csv) per keypoint and frame.

We also compute all unsupervised losses, where applicable, and store them (per keypoint and frame) in the following csvs:

  • predictions_pca_multiview_error.csv: pca multiview reprojection error between predictions and labeled keypoints

  • predictions_pca_singleview_error.csv: pca singleview reprojection error between predictions and labeled keypoints