EchoNet-Pediatric

# EchoNet-Pediatric Official EchoNet-Pediatric A4C and PSAX echocardiography videos and labels, plus EchoFlow `avae-4f4` latent archives used by this project. Official website: https://echonet.github.io/pediatric/ ## Contents | Path | Description | |---|---| | `Dataset.tar.gz` | Official A4C + PSAX videos, FileList/VolumeTracings CSVs | | `EchoFlow_20260605_latent.tar.gz` | EchoFlow `avae-4f4` latents (2026-06-05 encoding) | | `EchoFlow_20260615_echosyn_avae-4f4.tar.gz` | EchoFlow `avae-4f4` latents (2026-06-15 `echoflow-standard` encoding) | ### Official split sizes (`FileList_new.csv`) | View | TRAIN | VAL | TEST | Total | |---|---:|---:|---:|---:| | A4C | 2580 | 336 | 368 | 3284 | | PSAX | 3559 | 448 | 519 | 4526 | | **Total** | **6139** | **784** | **887** | **7810** | `FileList.csv` uses 10-fold cross-validation indices (`Split` 0–9). `FileList_new.csv` maps folds to `TRAIN` / `VAL` / `TEST` and adds frame metadata (`FrameHeight`, `FrameWidth`, `FPS`, `NumberOfFrames`). ## Dataset archive layout After extracting `Dataset.tar.gz`: ```text Dataset/ A4C/ Videos/ # 3284 .avi files FileList.csv FileList_new.csv VolumeTracings.csv VolumeTracings_new.csv PSAX/ Videos/ # 4526 .avi files FileList.csv FileList_new.csv VolumeTracings.csv VolumeTracings_new.csv ``` ```bash tar -xzf Dataset.tar.gz find Dataset/A4C/Videos -name '*.avi' | wc -l # expect 3284 find Dataset/PSAX/Videos -name '*.avi' | wc -l # expect 4526 ``` ## EchoFlow AVAE latent archives Both latent archives use EchoFlow VAE `avae-4f4`. Latent tensors are `torch.Tensor` with shape `[T, 4, 28, 28]` (`float32`). Each archive contains `ped_a4c/` and `ped_psax/` subdirectories with `Latents/`, `FileList.csv`, and `conversion_config.json`. | Archive | A4C cases | PSAX cases | Total | |---|---:|---:|---:| | `EchoFlow_20260605_latent.tar.gz` | 3284 | 4526 | 7810 | | `EchoFlow_20260615_echosyn_avae-4f4.tar.gz` | 3284 | 4526 | 7810 | The two archives cover the same cases but come from different encoding runs and are **not** numerically identical. - Prefer `EchoFlow_20260615_echosyn_avae-4f4.tar.gz` for new EchoFlow-standard experiments (`encoding_mode: echoflow-standard`). - Keep `EchoFlow_20260605_latent.tar.gz` for reproducing earlier 2026-06-05 results. ```bash tar -xzf EchoFlow_20260615_echosyn_avae-4f4.tar.gz tar -xzf EchoFlow_20260605_latent.tar.gz ``` ## Download Pin a full 40-hex ModelScope dataset revision when deploying: ```bash modelscope download \ --dataset LuminousXHJ/EchoNet-Pediatric \ --revision <40-hex-commit> \ --local_dir /path/to/EchoNet-Pediatric ``` ## Not included EchoNet-Synthetic reproduction artifacts (`EchoNet-Synthetic_20260604_*`), local symlinks, and staging caches are intentionally omitted from this dataset repository.