CSI classification tasks Tutorial

# !pip install pysensing

In this tutorial, we will be implementing codes for CSI classification tasks, including Human activity recognition and Human identity detection

import sys
sys.path.append('../..')
import pysensing.csi.dataset.get_dataloader as get_dataloader
import pysensing.csi.model.get_model as get_model
import pysensing.csi.inference.predict as predict
import pysensing.csi.inference.train as train
import pysensing.csi.inference.embedding as embedding
import torch

Load the data

# Human action recognition dataset:

# UT-HAR
# CSI size : 1 x 250 x 90
# number of classes : 7
# classes : lie down, fall, walk, pickup, run, sit down, stand up
# train number : 3977
# test number : 996

# NTU-HAR
# CSI size : 3 x 114 x 500
# number of classes : 6
# classes : box, circle, clean, fall, run, walk
# train number : 936
# test number : 264

# Widar
# BVP size : 22 x 20 x 20
# number of classes : 22
# classes :
# Push&Pull, Sweep, Clap, Slide, Draw-N(H), Draw-O(H),Draw-Rectangle(H),
# Draw-Triangle(H), Draw-Zigzag(H), Draw-Zigzag(V), Draw-N(V), Draw-O(V), Draw-1,
# Draw-2, Draw-3, Draw-4, Draw-5, Draw-6, Draw-7, Draw-8, Draw-9, Draw-10
# train number : 34926
# test number : 8726

# Human identity detection dataset:

# NTU-HumanID
# CSI size : 3 x 114 x 500
# number of classes : 14
# classes : gaits of 14 subjects
# train number : 546
# test number : 294
# Examples of NTU-Fi data


train_loader, test_loader = get_dataloader.load_classification_dataset('UT_HAR', batch_size=1)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
for data in train_loader:

    csi, label = data
    csi = csi.to(device)
    label = label.type(torch.LongTensor).to(device)
    print('data:', csi)
    print('label:', label)
    break

/data1/msc/zyj/yunjiao_csi/1028/yunjiao_csi/tutorials/csi_source/data/csi
using dataset: UT-HAR DATA
data: tensor([[[0.5545, 0.5545, 0.6269,  ..., 0.7542, 0.7898, 0.7559],
         [0.5545, 0.5545, 0.6269,  ..., 0.7542, 0.7898, 0.7559],
         [0.5545, 0.5545, 0.6269,  ..., 0.7542, 0.7898, 0.7559],
         ...,
         [0.5694, 0.6034, 0.6449,  ..., 0.7600, 0.7859, 0.7552],
         [0.5810, 0.6046, 0.6436,  ..., 0.7526, 0.7971, 0.7458],
         [0.5810, 0.6046, 0.6436,  ..., 0.7526, 0.7971, 0.7458]]],
       device='cuda:0')
label: tensor([1], device='cuda:0')

Load the model

Model zoo: MLP LeNet ResNet RNN GRU LSTM BiLSTM CNN+GRU ViT

model = get_model.load_har_model('UT_HAR', 'MLP')
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = get_model.load_pretrain(model, 'UT_HAR', 'MLP', device=device)
print(model)

har_MLP(
  (fc): Sequential(
    (0): Linear(in_features=22500, out_features=1024, bias=True)
    (1): ReLU()
    (2): Linear(in_features=1024, out_features=128, bias=True)
    (3): ReLU()
    (4): Linear(in_features=128, out_features=7, bias=True)
  )
)

Model train

criterion = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)
epoch_num = 1

train.har_train(train_loader, model, epoch_num, optimizer, criterion, device)

Epoch:1, Loss:1.901668259

Model inference

model = get_model.load_pretrain(model, 'UT_HAR', 'MLP', device=device)
output = predict.har_predict(csi, 'UT_HAR', model, device).type(torch.FloatTensor).to(device)
print("output:", output)

output: tensor([[-11.4981,   7.3044,  -7.5377,  -2.6977,  -4.2921,   0.1082,  -0.2130]],
       device='cuda:0')

Evaluate loss

criterion = torch.nn.CrossEntropyLoss()
loss = criterion(output, label)
print(loss)

tensor(0.0013, device='cuda:0')

Generate embeddings

csi_embedding = embedding.har_csi_embedding(csi, 'UT_HAR', model, device)
print('csi_embedding: ', csi_embedding)

csi_embedding:  tensor([[ 0.0000, 11.0925,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
          0.0000,  6.3804,  9.9213,  0.0000,  0.0000, 29.4593,  0.0000,  2.3160,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  9.4756,  0.0000, 12.5232,
         17.5423,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000, 11.4202,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  8.0701,  0.0000, 15.2518,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0420,
          0.0000,  0.0000,  0.0000,  9.9100,  0.0000,  0.0000,  0.0000,  0.0000,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
          0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  4.6838,  0.0000,  0.0000,
          5.1499,  0.0000,  0.0000,  7.9135,  0.0000,  0.0000,  0.0000,  0.0000,
          0.0000,  0.0000,  3.1761,  0.0000, 15.9863,  0.0000,  0.0000,  0.0000,
          0.0000, 17.0392,  4.3671,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000]],
       device='cuda:0', grad_fn=<ReluBackward0>)

And that’s it. We’re done with our CSI humna activity recognition and human identity detection tutorials. Thanks for reading.