Faster-Mean-Shift-Euc

Faster Mean-shift algorithm with Euclidean Distance Metrics. The algorithm is based on GPU acceleration, which can achieve satisfactory speedup with optimized GPU memory consumption. Here is a brief introduction to how to run it. And the details of our algorithm, please refer to our paper1, paper2.

The Cosine Metrics version is provided in another repository Faster-Mean-Shift.

Environment

Win10

VS2019

Anacoda 2020.11

The packages requirement please see requirements.txt

Please make sure to use a compatible pytorch-gpu version.

Example

Using our algorithm is similar to calling the meanshift in sklearn. An example of how to run the algorithm is given in main.py：

# Generate a blob dataset.
n_samples = 100000
blobs = datasets.make_blobs(n_samples=n_samples, random_state=9)

# Normalize dataset for easier parameter selection
X, y = blobs
X = StandardScaler().fit_transform(X)

# Estimate bandwidth for mean shift(Select 1000 points)
bandwidth = cluster.estimate_bandwidth(X[0:999])
bandwidth_gpu = 2*bandwidth/(X.max()-X.min())

# Obtain results
ms = MeanShiftEuc(bandwidth=bandwidth_gpu, cluster_all=True, GPU=True)
ms.fit(X)
labels  =  ms.labels_

Here is a helpful Q&A for image segmentation and bandwidth setting: #1

If you encounter any problem or find a bug during using, you are very welcome to contact me by (Mengyang.Zhao.TH@dartmouth.edu). If you use this code for your research, please cite our paper1, paper2. Thanks!