bit-plane-snn

Notes: work accepted for publication at IEEE Access

The official repository of "Improvement of Spiking Neural Network with Bit Planes and Color Models", contain the mentioned hybrid encoder.

Paper

Quickstart

This is a quickstart guide on how to use our encoder as a package

Step 1: Environment setup and repo download

To setup the environment testing with this encoder, you will need Pytorch and SpikingJelly. We suggest using conda environment with:

$ conda create -n env python=3.12.2
$ conda install pytorch=2.3.0 torchvision=0.18.0 torchaudio=2.3.0 pytorch-cuda=12.1 -c pytorch -c nvidia #As latest pytorch conda guide, change cuda version suitable to your case.
$ pip install spikingjelly
$ pip install git+https://github.com/luutn2002/bit-plane-snn.git

or clone and modify locally:

$ git clone https://github.com/luutn2002/bit-plane-snn.git

Step 2: Import and usage

To use the encoder, we can simply import and encode with:

from bit_plane_encoder.transform import PILToTensor
from bit_plane_encoder import MixedEncoder #Bit plane encode with rate coding
from PIL import Image
import requests
import torchvision

#Load image to PIL image
img = Image.open(requests.get("https://dummyimage.com/500x400/ff6699/000", stream=True).raw)

#Transform PIL images to tensor
transform = torchvision.transforms.Compose([
        PILToTensor(), #Conver PIL to unnormalized tensor.
        torchvision.transforms.Resize((224, 224)),
    ])

tensor = transform(img)
encoder = MixedEncoder(T = 10, #Time step of rate coding 
                    color_model ="rgb") #Selected color model information to be used, default to rgb.
encoded_tensor = encoder(tensor)

To only use the bit plane encoder, you can use:

from bit_plane_encoder.transform import PILToTensor
from bit_plane_encoder import BitplaneColorEncoder #Bit plane encode only
from PIL import Image
import requests
import torchvision

#Load image to PIL image
img = Image.open(requests.get("https://dummyimage.com/500x400/ff6699/000", stream=True).raw)

#Transform PIL images to tensor
transform = torchvision.transforms.Compose([
        PILToTensor(), #Conver PIL to unnormalized tensor.
        torchvision.transforms.Resize((224, 224)),
    ])

tensor = transform(img)
encoder = BitplaneColorEncoder(color_model ="rgb") #Selected color model information to be used, default to rgb.
encoded_tensor = encoder(tensor)

Current supported color models include: RGB, CMY, YCbCr, HSL, HSV, LAB, XYZ

color_models = ["rgb", "cmy", "ycbcr", "hsl", "hsv", "lab", "xyz"]
encoder = BitplaneColorEncoder(color_model = color_models[0])

Acknowledgement

Special thanks to SpikingJelly for their guides to SNN and Kornia for color conversion algorithm reference.

Citation

@ARTICLE{11261679,
  author={Luu, Nhan T. and Luu, Duong T. and Nam, Pham Ngoc and Thang, Truong Cong},
  journal={IEEE Access}, 
  title={Improvement of Spiking Neural Network With Bit Planes And Color Models}, 
  year={2025},
  volume={},
  number={},
  pages={1-1},
  keywords={Neurons;Image color analysis;Training;Spiking neural networks;Computational modeling;Tensors;Encoding;Biological system modeling;Deep learning;Membrane potentials;Spiking neural network;spike coding;image classification;bit planes;color models},
  doi={10.1109/ACCESS.2025.3635297}}