3D shape analysis using deep learning
Project description
Cellshape-cluster is an easy-to-use tool to analyse the cluster cells by their shape using deep learning and, in particular, deep-embedded-clustering. The tool provides the ability to train popular graph-based or convolutional autoencoders on point cloud or voxel data of 3D single cell masks as well as providing pre-trained networks for inference.
To install
pip install cellshape-cluster
Usage
Basic usage:
import torch
from cellshape_cloud import CloudAutoEncoder
from cellshape_cluster import DeepEmbeddedClustering
autoencoder = CloudAutoEncoder(
num_features=128,
k=20,
encoder_type="dgcnn"
)
model = DeepEmbeddedClustering(autoencoder=autoencoder,
num_clusters=10,
alpha=1.0)
points = torch.randn(1, 2048, 3)
recon, features, clusters = model(points)
To load a trained graph-based autoencoder and perform deep embedded clustering:
import torch
from torch.utils.data import DataLoader
import cellshape_cloud as cloud
import cellshape_cluster as cluster
from cellshape_cloud.vendor.chamfer_distance import ChamferDistance
dataset_dir = "path/to/pointcloud/dataset/"
autoencoder_model = "path/to/autoencoder/model.pt"
num_features = 128
k = 20
encoder_type = "dgcnn"
num_clusters = 10
num_epochs = 1
learning_rate = 0.00001
gamma = 1
divergence_tolerance = 0.01
output_dir = "path/to/output/"
autoencoder = CloudAutoEncoder(
num_features=128,
k=20,
encoder_type="dgcnn"
)
checkpoint = torch.load(autoencoder_model)
autoencoder.load_state_dict(checkpoint['model_state_dict']
model = DeepEmbeddedClustering(autoencoder=autoencoder,
num_clusters=10,
alpha=1.0)
dataset = cloud.PointCloudDataset(dataset_dir)
dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=False) # it is very important that shuffle=False here!
dataloader_inf = DataLoader(dataset, batch_size=1, shuffle=False) # it is very important that batch_size=1 and shuffle=False here!
optimizer = torch.optim.Adam(
model.parameters(),
lr=learning_rate * 16 / batch_size,
betas=(0.9, 0.999),
weight_decay=1e-6,
)
reconstruction_criterion = ChamferDistance()
cluster_criterion = nn.KLDivLoss(reduction="sum")
train(
model,
dataloader,
dataloader_inf,
num_epochs,
optimizer,
reconstruction_criterion,
cluster_criterion,
update_interval,
gamma,
divergence_tolerance,
output_dir
)
Parameters
autoencoder
: CloudAutoEncoder or VoxelAutoEncoder.
Instance of autoencoder class from cellshape-cloud or cellshape-voxelnum_clusters
: int.
The number of clusters to use in deep embedded clustering algorithm.alpha
: float.
Degrees of freedom for the Student's t-distribution. Xie et al. (ICML, 2016) let alpha=1 for all experiments.
For developers
- Fork the repository
- Clone your fork
git clone https://github.com/USERNAME/cellshape-cluster
- Install an editable version (
-e
) with the development requirements (dev
)
cd cellshape-cluster
pip install -e .[dev]
- To install pre-commit hooks to ensure formatting is correct:
pre-commit install
- To release a new version:
Firstly, update the version with bump2version (bump2version patch
,
bump2version minor
or bump2version major
). This will increment the
package version (to a release candidate - e.g. 0.0.1rc0
) and tag the
commit. Push this tag to GitHub to run the deployment workflow:
git push --follow-tags
Once the release candidate has been tested, the release version can be created with:
bump2version release
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for cellshape-cluster-0.0.14rc0.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | f7e99884f5fd8adcffbad5b598e94955f70d42793dd4abd5f8db7f422786d03f |
|
MD5 | 316b7ed7d61b91dff6d6bf66cf6e045f |
|
BLAKE2b-256 | 56913f3149013064b0673060123c3d011fae5f6a876b8902c6c39eaf0c0c4b16 |
Hashes for cellshape_cluster-0.0.14rc0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 11f2ce0877635a462b1a8eeec5e6d2e15f5692178bb910b6654b3b931ebc953e |
|
MD5 | 183b8b2c63bb3f062b8516bf5a81844b |
|
BLAKE2b-256 | 930f892146b9ba4fdca91e17441a38397a6ce340facf895cd22da18145e4551b |