### Install cellSAM Library

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/reproducibility.md

Install the cellSAM library from the parent directory after activating the virtual environment.

```bash
pip install ..
```

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

### Manual Installation of Napari Dependencies

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/napari.md

Manually install napari and PySide2 if you need finer-grained control over Qt bindings.

```bash
pip install napari pyside2 magicgui
```

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

### Clone Repository

Source: https://github.com/vanvalenlab/cellsam/blob/master/cellSAM/AnchorDETR/README.md

Clone the AnchorDETR repository locally to get started.

```bash
git clone https://github.com/megvii-research/AnchorDETR.git
```

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

### Install CellSAM

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/index.md

Install the CellSAM package using pip. This command clones the repository and installs the package.

```bash
pip install git+https://github.com/vanvalenlab/cellSAM.git
```

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

### Install CellSAM with Napari Dependencies

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/napari.md

Install cellSAM with optional napari dependencies. This is the simplest way to ensure all necessary packages are installed.

```bash
pip install .[napari]
```

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

### Install CellSAM from GitHub with Napari

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/napari.md

Install cellSAM directly from GitHub, including napari dependencies, for the latest version.

```bash
pip install "cellSAM[napari] @ git+https://github.com/vanvalenlab/cellSAM@master"
```

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

### Install Evaluation Dependencies

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/reproducibility.md

Install the specific requirements for the evaluation suite using the provided requirements.txt file. This may downgrade some existing dependencies.

```bash
pip install -r requirements.txt
```

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

### Unpack Evaluation Dataset with Parallel Decompression

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/reproducibility.md

Unpack the downloaded dataset archive to a specified location, utilizing parallel decompression for efficiency. This example uses 8 threads and unpacks to the /data directory.

```bash
tar --use-compress-program="unpigz -p 8" -xf $HOME/.deepcell/datasets/cellsam-data_v1.2.tar.gz -C /data
```

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

### Run the CellSAM Napari Plugin

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/napari.md

Launch the napari application with the cellSAM plugin enabled from the command line.

```bash
cellsam napari
```

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

### Run All Evaluations

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/reproducibility.md

Execute the entire evaluation suite using the all_runs.sh script. Ensure model and dataset locations are correctly configured at the top of the script.

```bash
./all_runs.sh
```

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

### Create and Activate Python Virtual Environment

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/reproducibility.md

Use Python's built-in venv module to create a new virtual environment for evaluation. Ensure you replace 'XX' with your desired Python version.

```bash
python3.XX -m venv cs-eval-env
source cs-eval-env/bin/activate
```

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

### Download Pretrained Model Weights

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/reproducibility.md

Download and unpack the latest version of the pretrained model weights using the get_model function from the cellSAM library. Specify a version using `version=` for specific model weights.

```python
from cellSAM import get_model
get_model();
```

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

### Download Training Data

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/API-key.md

Use this function to download the training dataset, which is approximately 14GB. Ensure you have sufficient disk space and network bandwidth before proceeding. The compressed dataset will be stored in $HOME/.deepcell/data/cellsam/.

```python
from cellSAM import download_training_data

download_training_data()
```

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

### Visualize Bounding Box for Segmentation

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/tutorial.md

Defines a bounding box and visualizes it on the input image. This is used to demonstrate how to prompt CellSAM for specific cell segmentation.

```python
# Here's the cell we want to segment!
box = [290, 365, 60, 38] # x, y, w, h

rect = patches.Rectangle(
    (box[0], box[1]), box[2], box[3], linewidth=1, edgecolor='r', facecolor='none'
)

plt.figure(figsize=(5, 5))
plt.imshow(img, cmap='gray')
plt.gca().add_patch(rect);
```

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

### Train AnchorDETR (8 GPUs)

Source: https://github.com/vanvalenlab/cellsam/blob/master/cellSAM/AnchorDETR/README.md

Train the AnchorDETR model on a single node using 8 GPUs. Ensure the COCO dataset path is correctly specified.

```python
python -m torch.distributed.launch --nproc_per_node=8 --use_env main.py  --coco_path /path/to/coco 
```

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

### Initialize CellSAM Predictor and Segment Specific Cell

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/tutorial.md

Initializes the CellSAM model and preprocesses the image. It then uses the provided bounding box to predict a mask for a specific cell.

```python
predictor = get_model()

# We can pass the bounding boxes to the model prediction function
x1, y1, w, h = box

# Bounding-box prompts should have format (x1, x2, y1, y2), where (x1, y1) is
# the lower-left corner of the box and (x2, y2) is the upper right
x2, y2 = x1 + w, y1 + h

# Preprocess the image
im = format_image_shape(img)
im = normalize_image(im)
im = np.transpose(im, (2, 0, 1))
im = np.expand_dims(im, axis=0)
pred_mask, *_ = predictor.predict(im, boxes_per_heatmap=[[[x1, y1, x2, y2]]])
```

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

### Import CellSAM Libraries

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/tutorial.md

Imports necessary libraries for using CellSAM, including image processing and visualization tools.

```python
import imageio.v3 as iio
import numpy as np
import scipy as sp
import matplotlib.pyplot as plt
import matplotlib.patches as patches

from cellSAM import cellsam_pipeline, get_model
from cellSAM.utils import format_image_shape, normalize_image
```

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

### Download Evaluation Dataset

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/reproducibility.md

Initiate the download of the compressed evaluation dataset archive. The data will be saved to $HOME/.deepcell/datasets/.

```python
from cellSAM import download_training_data
download_training_data()
```

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

### Run CellSAM Inference Pipeline

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/tutorial.md

Loads an image and runs the `cellsam_pipeline` to predict masks for all cells. This method automatically finds bounding boxes if none are provided. Performance is significantly better on a GPU.

```python
# Load sample data and trim padding
img = iio.imread("../sample_imgs/YeaZ.png")[8:-7, 8:-7]

# Run inference pipeline
mask = cellsam_pipeline(
    img, use_wsi=False, low_contrast_enhancement=False, gauge_cell_size=False
)

# Visualize results
plt.imshow(mask);
```

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

### Evaluate AnchorDETR (8 GPUs)

Source: https://github.com/vanvalenlab/cellsam/blob/master/cellSAM/AnchorDETR/README.md

Evaluate the trained AnchorDETR model on a single node with 8 GPUs. Specify the evaluation flag, COCO path, and the path to the pre-trained checkpoint.

```python
python -m torch.distributed.launch --nproc_per_node=8 --use_env main.py --eval --coco_path /path/to/coco --resume /path/to/checkpoint.pth 
```

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

### Evaluate AnchorDETR (Single GPU)

Source: https://github.com/vanvalenlab/cellsam/blob/master/cellSAM/AnchorDETR/README.md

Evaluate the trained AnchorDETR model using a single GPU. Provide the COCO dataset path and the checkpoint path.

```python
python main.py --eval --coco_path /path/to/coco --resume /path/to/checkpoint.pth
```

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

### Perform Cell Segmentation

Source: https://github.com/vanvalenlab/cellsam/blob/master/README.md

Load a sample image and perform cellular segmentation using the `segment_cellular_image` function. Ensure you have a CUDA-enabled device for optimal performance.

```python
import numpy as np
from cellSAM import segment_cellular_image
img = np.load("sample_imgs/yeaz.npy")
mask, _, _ = segment_cellular_image(img, device='cuda')
```

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

### Set DeepCell Access Token Environment Variable

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/API-key.md

Set the DEEPCELL_ACCESS_TOKEN environment variable with your token obtained from users.deepcell.org. This allows automatic access to DeepCell models and data upon login when added to your shell configuration.

```bash
export DEEPCELL_ACCESS_TOKEN=<token-from-users.deepcell.org>
```

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

### CellSAM Citation

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/index.md

BibTeX entry for citing the CellSAM foundation model in academic work.

```latex
@article{israel2023foundation,
title={A Foundation Model for Cell Segmentation},
author={Israel, Uriah and Marks, Markus and Dilip, Rohit and Li, Qilin and Schwartz, Morgan and Pradhan, Elora and Pao, Edward and Li, Shenyi and Pearson-Goulart, Alexander and Perona, Pietro and others},
journal={bioRxiv},
publisher={Cold Spring Harbor Laboratory Preprints},
doi = {10.1101/2023.11.17.567630},
}
```

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

### Anchor DETR Citation

Source: https://github.com/vanvalenlab/cellsam/blob/master/cellSAM/AnchorDETR/README.md

Cite the Anchor DETR paper in your research if this project is useful. Includes author, title, booktitle, volume, number, pages, and year.

```bibtex
@inproceedings{wang2022anchor,
  title={Anchor detr: Query design for transformer-based detector},
  author={Wang, Yingming and Zhang, Xiangyu and Yang, Tong and Sun, Jian},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={36},
  number={3},
  pages={2567--2575},
  year={2022}
}
```

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

### Visualize Predicted Cell Mask

Source: https://github.com/vanvalenlab/cellsam/blob/master/docs/tutorial.md

Superimposes the predicted cell mask as an edge onto the original image for clear visualization. This helps in verifying the segmentation accuracy.

```python
# Use several iterations to make the mask edge visible when plotting
dilated_mask = sp.ndimage.binary_dilation(pred_mask > 0, iterations=5)
edges = (dilated_mask > pred_mask).astype(np.uint8)

# Plot the image
plt.imshow(img, cmap="gray")

# And the outlines from the mask
plt.imshow(255 * edges, cmap="Reds", alpha=edges);
```

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

### CellSAM Citation

Source: https://github.com/vanvalenlab/cellsam/blob/master/README.md

BibTeX entry for citing the CellSAM paper. Please cite this work when using CellSAM in your research.

```bibtex
@article{israel2023foundation,
  title={A Foundation Model for Cell Segmentation},
  author={Israel, Uriah and Marks, Markus and Dilip, Rohit and Li, Qilin and Schwartz, Morgan and Pradhan, Elora and Pao, Edward and Li, Shenyi and Pearson-Goulart, Alexander and Perona, Pietro and others},
  journal={bioRxiv},
  publisher={Cold Spring Harbor Laboratory Preprints},
  doi = {10.1101/2023.11.17.567630},
}
```

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