### Setup Conda Environment and Install PyTorch

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/README.md

This snippet shows how to create a conda environment from an 'environment.yml' file, activate it, and then install specific versions of PyTorch and Torchvision for CUDA 11.8.

```bash
conda env create -f environment.yml
conda activate hovernext
pip install torch==2.1.1 torchvision==0.16.1 --index-url https://download.pytorch.org/whl/cu118
```

--------------------------------

### Run Docker/Singularity Container

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/README.md

Command to download the Singularity image for HoVer-Next inference.

```bash
singularity run --nv hover_next.sif
```

--------------------------------

### Run Inference on TCGA Sample

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/sample_analysis.ipynb

Initiates the main inference process with specified parameters. Ensure the 'main' module is correctly imported.

```python
import zarr
import numpy as np
import json
from main import main

'''
Run inference on a small sample image from TCGA
'''
params = {

}

main(params)
```

--------------------------------

### Load Instance Map from Zarr

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/sample_analysis.ipynb

Opens a Zarr file for interaction with instance map data. Supports ROI selection or full array loading.

```python
'''
Instance map: 2D full-size matrix where each pixels value corresponds to the associated instance (value>0) or background (value=0)
'''

# open: file-like interaction with zarr-array
instance_map = zarr.open("pinst_pp.zip", mode="r")
# selecting a ROI will yield a numpy array
roi = instance_map[10000:20000,10000:20000]
# or with [:] to load the entire array
full_instance_map = instance_map[:]
# alternatively, use load, which will directly create a numpy array:
full_instance_map = zarr.load("pinst_pp.zip")
```

--------------------------------

### Run HoVer-NeXt Inference with Apptainer

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/README.md

Execute the HoVer-NeXt inference pipeline using Apptainer. Ensure your local directory is mounted to APPTAINER_BINDPATH. This command processes SVS images and saves results.

```bash
export APPTAINER_BINDPATH="/storage"
apptainer exec --nv /path-to-container/hover_next.sif \
    python3 /path-to-repo/main.py \
    --input "/path-to-wsi/*.svs" \
    --output_root "results/"
	--cp "lizard_convnextv2_large" \
    --tta 4 
```

--------------------------------

### Create Class Map using Lookup

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/sample_analysis.ipynb

Generates a class map by creating a lookup array from class information and applying it to the full instance map.

```python
# or alternatively create a lookup for the instance map to get a corresponding class map
pcls_list = np.array([0] + [v[0] for v in class_info.values()])
pcls_keys = np.array(["0"] + list(class_info.keys())).astype(int)
lookup = np.zeros(pcls_keys.max() + 1,dtype=np.uint8)
lookup[pcls_keys] = pcls_list
cls_map = lookup[full_instance_map]
```

--------------------------------

### Load Class Information and Create Centroid Array

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/sample_analysis.ipynb

Loads class information from a JSON file and constructs a NumPy array containing instance IDs, class IDs, and centroid coordinates.

```python
'''
Class dictionary: Lookup for the instance map, also contains centroid coordinates. If only centroid coordinates are of interest, you can skip loading the instance map.
'''

# load the dictionary
with open("class_inst.json","r") as f:
    class_info = json.load(f)
# create a centroid info array
centroid_array = np.array([[int(k),v[0],*v[1]] for k,v in class_info.items()])
# [instance_id, class_id, y, x]
```

--------------------------------

### Run WSI Inference

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/README.md

Command to perform inference on a single Whole Slide Image (WSI) using specified input, output, model checkpoint, and various inference parameters. Adjust parameters like '--inf_workers' and '--pp_workers' based on your machine's cores.

```bash
python3 main.py \
    --input "/path-to-wsi/wsi.svs" \
    --output_root "results/" \
    --cp "lizard_convnextv2_large" \
    --tta 4 \
    --inf_workers 16 \
    --pp_tiling 10 \
    --pp_workers 16
```

--------------------------------

### Run NPY or Image Inference

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/README.md

Command to perform inference on NPY or common image files. The output will be a ZARR array. Similar to WSI inference, optimize worker and tiling parameters for your system.

```bash
python3 main.py \
    --input "/path-to-file/file.npy" \
    --output_root "/results/" \
    --cp "lizard_convnextv2_large" \
    --tta 4 \
    --inf_workers 16 \
    --pp_tiling 10 \
    --pp_workers 16
```

--------------------------------

### HoVer-NeXt Citation

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/README.md

Citation for the HoVer-NeXt paper published in Medical Imaging with Deep Learning.

```bibtex
@inproceedings{baumann2024hover,
  title={HoVer-NeXt: A Fast Nuclei Segmentation and Classification Pipeline for Next Generation Histopathology},
  author={Baumann, Elias and Dislich, Bastian and Rumberger, Josef Lorenz and Nagtegaal, Iris D and Martinez, Maria Rodriguez and Zlobec, Inti},
  booktitle={Medical Imaging with Deep Learning},
  year={2024}
}
```

--------------------------------

### Panoptic Segmentation Citation

Source: https://github.com/digitalpathologybern/hover_next_inference/blob/main/README.md

Citation for the Panoptic segmentation paper.

```bibtex
@INPROCEEDINGS {"rumberger2022panoptic",
  author={Rumberger, Josef Lorenz and Baumann, Elias and Hirsch, Peter and Janowczyk, Andrew and Zlobec, Inti and Kainmueller, Dagmar},
  booktitle={2022 IEEE International Symposium on Biomedical Imaging Challenges (ISBIC)},
  title={Panoptic segmentation with highly imbalanced semantic labels},
  year={2022},
  pages={1-4},
  doi={10.1109/ISBIC56247.2022.9854551}}

```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.