# tmp.bonnet

**Repository Path**: tyloeng/tmp.bonnet

## Basic Information

- **Project Name**: tmp.bonnet
- **Description**: https://github.com/HanjiangZhu/Bonnet
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2026-02-21
- **Last Updated**: 2026-02-21

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# iTom's Notes

- environment: [env.sh](env.sh)
- download pretrained weight: [download.sh](download.sh)
- link Ribsegv2 data: [link.sh](link.sh)
- inference with pretrained model: [run_ribsegv2.sh](run_ribsegv2.sh)
    - my adapted inference code: [src/test.py](src/test.py)

## adaptations

Add [src/utils/bin_packing.py](src/utils/bin_packing.py):
this file is needed but missing in original repository.
I ask Claude to write it.
Not sure how this will affect the model performance.


For Ribsegv2 dataset:
- `Ribsegv2Dataset` in [src/data/ctdataset.py](src/data/ctdataset.py): mainly adapt the `create_sample` function, change the data loading.
- `Ribsegv2` in [src/data/data_utils.py](src/data/data_utils.py): meta-data & helper functions.

## notes

### 20 Feb 2026

Directly using the downloaded weights to test on Ribsegv2 gives very bad results.
I guess training on Ribsegv2 is needed to get reasonable results.

# Bonnet: Ultra-Fast Whole-Body Bone Segmentation from CT Scans

Bonnet is an ultra-fast whole-body bone segmentation pipeline for CT scans. It runs in seconds per scan on a single commodity GPU while maintaining reliable segmentation quality across different datasets.

## Train

1. Set dataset / output paths and other options in:

- `Bonnet/conf/config_eva.yaml`

2. Run:

```
python main.py
```

## Evaluate only
Model checkpoint (Hugging Face): [https://huggingface.co/hanjiangjiang123/Bonnet](https://huggingface.co/hanjiangjiang123/Bonnet)

1. Open:

- `Bonnet/conf/eval/eval_on_test.yaml`

2. Set:

```
eval_only: True
```

3. Run:

```
python main.py
```