---
language: en
license: lgpl-3.0
tags:
  - pytorch
  - torchscript
  - image-segmentation
  - particle-physics
  - neutrino-detectors
  - liquid-argon
datasets:
  - custom
library_name: pytorch
pipeline_tag: image-segmentation
model-index:
  - name: DNN ROI Models
    results:
      - task:
          type: image-segmentation
          name: Region of Interest Detection
        metrics:
          - type: iou
            value: N/A
            name: Intersection over Union
          - type: dice
            value: N/A
            name: Dice Coefficient
models:
  - icarus/moon-2025-08-25
  - pdhd/dikshant/mobileunet
  - pdhd/hokyeong/mobilenetv3
  - pdsp/unet
  - pdsp/nestedunet
  - sbnd/sbnd_data-v01_34_00
---

# DNN ROI Models

A collection of deep neural network models for region of interest (ROI) detection in LArTPC experiments, including ICARUS, ProtoDUNE-HD, ProtoDUNE-SP, and SBND.

```bash
├── icarus
│   └── moon-2025-08-25
│       ├── plane0_rand.ts
│       └── plane1_rand.ts
├── pdhd
│   ├── dikshant
│   │   ├── mobileunet_largedataset_fullimage.ts
│   │   ├── mobileunet_largedataset_rebin4.ts
│   │   ├── unet_largedataset_fullimage.ts
│   │   └── unet_largedataset_rebin4.ts
│   └── hokyeong
│       └── CP49_mobilenetv3.ts
├── pdsp
│   ├── pth-model # models in pytorch pickle format
│   │   ├── nestedunet-l23-cosmic500-e50.pth # input: loose, MP2, MP3
│   │   ├── unet-l23-cosmic500-e50.pth # input: loose, MP2, MP3
│   │   └── unet-lt-cosmic500-e50.pth # input: loose, tight
│   ├── ts-model-1.3 # TorchScript model saved using PyTorch 1.3
│   │   ├── nestedunet-l23-cosmic500-e50.ts
│   │   ├── unet-l23-cosmic500-e50.ts
│   │   └── unet-lt-cosmic500-e50.ts
│   └── ts-model-2.3 # TorchScript model saved using PyTorch 2.3
│       ├── nestedunet-l23-cosmic500-e50.ts
│       ├── unet-l23-cosmic500-e50.ts
│       └── unet-lt-cosmic500-e50.ts
├── README.md
└── sbnd
    └── sbnd_data-v01_34_00
        ├── plane0.ts
        └── plane1.ts
```