https://github.com/johnpertoft/spherical-hashing

A Tensorflow 2 implementation of Spherical Hashing
https://github.com/johnpertoft/spherical-hashing

Last synced: about 1 month ago
JSON representation

A Tensorflow 2 implementation of Spherical Hashing

• Host: GitHub
• URL: https://github.com/johnpertoft/spherical-hashing
• Owner: johnPertoft
• License: mit
• Created: 2020-09-03T07:29:18.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-09-08T15:34:42.000Z (over 4 years ago)
• Last Synced: 2024-12-02T23:35:18.209Z (2 months ago)
• Language: Python
• Homepage:
• Size: 13.7 KB
• Stars: 8
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Spherical Hashing

A Tensorflow 2 implementation of [Spherical Hashing](https://ieeexplore.ieee.org/document/6248024).

Spherical hashing is a way to compute a binary encoding of a feature vector while still maintaining

spatial coherence. This binary encoding can then be used for an approximate nearest neighbor 

solution since the compactness of this representation allows for faster neighbor search.

## Usage

```

from spherical_hashing import train_spherical_hashing

import tensorflow as tf

x_train = tf.random.uniform(shape=(1000, 512), minval=-10.0, maxval=10.0)

sph_model = train_spherical_hashing_model(x_train, n_bits=32)

x_test = tf.random.uniform(shape=(100, 512), minval=-10.0, maxval=10.0)

bits = sph_model(x_test, apply_pack_bits=True)

```

## Installation

```

pip install tf-spherical-hashing

```

## Development

### Run tests

```

./scripts/run-tests.sh

```

### Publish pip package

```

./scripts/publish.sh

```