Hand_sign_Pytorch.ipynb

{
  "cells": [
    {
      "metadata": {
        "_uuid": "22680c15f9e1abb40a9a991c644db5ee00c7b419"
      },
      "cell_type": "markdown",
      "source": "# Hand Sign Image Recognition using Pytorch Deep Neural Network  "
    },
    {
      "metadata": {
        "_uuid": "bc222a5477da00c32fbabdb12122ca811267ae50"
      },
      "cell_type": "markdown",
      "source": "In this notebook , I am building a Hand Sign Recognition system with Deep NN using Pytorch . Hand Sign means numeric signs (e.g. 0,1,2,3,4,5) using our hand . We have built a deep learning neural network to accomplish our target . We are using Pytorch deep learning framework . "
    },
    {
      "metadata": {
        "_uuid": "ce1aa0b19e75581e15a1c40eab6c8ea57462c55d",
        "_cell_guid": "0945dfb3-98d2-439c-a15a-2799bc1091a5",
        "trusted": true
      },
      "cell_type": "code",
      "source": "# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load in \n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\nimport h5py\n# Input data files are available in the \"../input/\" directory.\n# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory\n\nfrom subprocess import check_output\nprint(check_output([\"ls\", \"../input\"]).decode(\"utf8\"))\n\n# Any results you write to the current directory are saved as output.",
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": "hand-sign\nhand-sign-test\n\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "e5fe8f723a4f6483158be7dc97aff96b284f48bb"
      },
      "cell_type": "markdown",
      "source": "The Hand sign data set that we will use is available in .h5 format . So, we use h5py package we load the data . Then we convert the data in numpy array . Each image is a 3 channel image with 64 * 64  dimension. There are 1080 images in the train data and 120 in test data. "
    },
    {
      "metadata": {
        "collapsed": true,
        "_uuid": "72e4a047f2f23efb35e934d655add68e0795baa9",
        "_cell_guid": "55c80a90-d1bb-4039-a134-50236ce3ae97",
        "trusted": true
      },
      "cell_type": "code",
      "source": "def load_dataset():\n    train_dataset = h5py.File('../input/hand-sign/train_signs.h5', \"r\")\n    train_set_x_orig = np.array(train_dataset[\"train_set_x\"][:]) # your train set features\n    train_set_y_orig = np.array(train_dataset[\"train_set_y\"][:]) # your train set labels\n\n    test_dataset = h5py.File('../input/hand-sign-test/test_signs.h5', \"r\")\n    test_set_x_orig = np.array(test_dataset[\"test_set_x\"][:]) # your test set features\n    test_set_y_orig = np.array(test_dataset[\"test_set_y\"][:]) # your test set labels\n\n    #classes = np.array(test_dataset[\"list_classes\"][:]) # the list of classes\n    \n    train_set_y_orig = train_set_y_orig.reshape((1, train_set_y_orig.shape[0]))\n    test_set_y_orig = test_set_y_orig.reshape((1, test_set_y_orig.shape[0]))\n    \n    return train_set_x_orig, train_set_y_orig,test_set_x_orig,test_set_y_orig",
      "execution_count": 5,
      "outputs": []
    },
    {
      "metadata": {
        "_uuid": "a17e74006ba619ddd23a511e07e5484dedad0efb"
      },
      "cell_type": "markdown",
      "source": "Load the data and have quick look at rhe shape of the data."
    },
    {
      "metadata": {
        "_uuid": "720adec529db3ddce1cc215d610c26c78968c975",
        "_cell_guid": "e578bd15-aa1f-4ab9-827d-ad39ad6d14c9",
        "trusted": true
      },
      "cell_type": "code",
      "source": "train_x,train_y,test_x,test_y = load_dataset()\ntrain_x.shape,train_y.shape,test_x.shape,test_y.shape",
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 7,
          "data": {
            "text/plain": "((1080, 64, 64, 3), (1, 1080), (120, 64, 64, 3), (1, 120))"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "1e3360300ed9762f6f9ea180a87a8194b8d3ad9a"
      },
      "cell_type": "markdown",
      "source": "Just to have a look at how the images look like we use matplotlib.pyplot.imshow to display 2 imges from training data and 2 from test data."
    },
    {
      "metadata": {
        "_uuid": "4f785679fe835d61f0ebfd8f8921239731446f05",
        "_cell_guid": "800899f0-706b-45b3-b1e4-83ce58f41847",
        "trusted": true
      },
      "cell_type": "code",
      "source": "import matplotlib.pyplot as plt\nplt.subplots(2,2)\nplt.subplots_adjust(top=0.92, bottom=0.08, left=0.10, right=0.95, hspace=0.45,\n                    wspace=0.45)\n\nplt.subplot(2,2,1)\nplt.title('train_x[5] label : 4')\nplt.imshow(train_x[5])\nplt.subplot(2,2,2)\nplt.title('train_x[10] label : 2')\nplt.imshow(train_x[10])\n\nplt.subplot(2,2,3)\nplt.title('test_x[5] label : 0')\nplt.imshow(test_x[5])\nplt.subplot(2,2,4)\nplt.title('test_x[10] label : 5')\nplt.imshow(test_x[10])\n",
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 33,
          "data": {
            "text/plain": "<matplotlib.image.AxesImage at 0x7f245af794e0>"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": "<matplotlib.figure.Figure at 0x7f24839325c0>",
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWcAAAEhCAYAAACjuWH8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "1d639a6cd60c58b952e9380733ed75e1722bd130"
      },
      "cell_type": "markdown",
      "source": "Importing all the torch libraries needed to bulid a Deep CNN.   "
    },
    {
      "metadata": {
        "collapsed": true,
        "_uuid": "56f3057480dc3fb61a80344ac50e893bd73f20d7",
        "_cell_guid": "bb3fcfdb-583d-40ef-9b7b-e22e4c83473a",
        "trusted": true
      },
      "cell_type": "code",
      "source": "import torch\nimport torch.nn as nn\nimport torch.utils.data as data\nimport torchvision.datasets as dsets\nimport torchvision.transforms as transforms\nfrom torch.autograd import Variable",
      "execution_count": 13,
      "outputs": []
    },
    {
      "metadata": {
        "_uuid": "7af08202fbc9033b2ca278be7a86904e2eaca5c9"
      },
      "cell_type": "markdown",
      "source": "In pytorch to use a data for training a deep nn model we need to build a class for that data . Then we create an data object by making an instance of the class and then using Dataloader to create mini random batches of data of batch size 64. This is important because we normally don't use the whole data for training the model at each iteration.  "
    },
    {
      "metadata": {
        "collapsed": true,
        "_uuid": "b24bb2d8355fcbf4d8bcc70ffa54a899b4fad2b3",
        "_cell_guid": "96d356d9-266e-4584-8413-18015e8667bb",
        "trusted": true
      },
      "cell_type": "code",
      "source": "class HandSign(data.Dataset):\n    def __init__(self):\n        train_X = train_x.reshape(1080,3,64,64)\n        train_Y = np.squeeze(train_y).tolist()\n        self.datalist = train_X\n        self.labellist = train_Y\n\n\n    def __getitem__(self, index):\n        return torch.Tensor(self.datalist[index].astype(float)), self.labellist[index]\n\n    def __len__(self):\n        return self.datalist.shape[0]\n\ntrain_data = HandSign()\ntrain_loader = torch.utils.data.DataLoader(dataset=train_data,\n                                           batch_size=64, \n                                           shuffle=True,\n                                           num_workers=2)",
      "execution_count": 14,
      "outputs": []
    },
    {
      "metadata": {
        "_uuid": "5cefc70ad33d13bade6f82876dd36146eeb6d5cd"
      },
      "cell_type": "markdown",
      "source": "Now create a nn.Module class CNN_sign where we define the model and the forward propagation approach . As it can be seen we  have taken two layer of (Conv2d , BatchNorm2d ,ReLU ,MaxPool2d ) . Then  we have add atleast one fully connected layer usinf nn.Linear . But we add 2 fully connected layer to have a better precision . And finally  I am adding one BatchNorm1d layer because it prevents the output from going beyong -1 to 1. "
    },
    {
      "metadata": {
        "_uuid": "e3719c37850d1ddaa8bf83964c1069ab033e50d9"
      },
      "cell_type": "markdown",
      "source": "I tried different combination of nn models and I finally found this one the best performing  so far. "
    },
    {
      "metadata": {
        "collapsed": true,
        "_uuid": "a9f57b1638ea2b6b897183303e827c66cae16fad",
        "_cell_guid": "0d0a740f-9e44-4d42-9ada-d62cac288dbe",
        "trusted": true
      },
      "cell_type": "code",
      "source": "class CNN_sign(nn.Module):\n    def __init__(self):\n        super(CNN_sign,self).__init__()\n        self.layer1 = nn.Sequential(\n            nn.Conv2d(3, 10, kernel_size=5),\n            nn.BatchNorm2d(10),\n            nn.ReLU(),\n            nn.MaxPool2d(2),\n            nn.Conv2d(10, 10, kernel_size=5),\n            nn.BatchNorm2d(10),\n            nn.ReLU(),\n            nn.MaxPool2d(2)\n        )\n        self.fc1 = nn.Linear(13*13*10, 100)\n        self.fc2 = nn.BatchNorm1d(100)\n        self.fc3 = nn.Linear(100,6)\n        \n    def forward(self, x):\n        out = self.layer1(x)\n        out = out.view(out.size(0),-1)\n        out = self.fc1(out)\n        out = self.fc2(out)\n        out = self.fc3(out)\n        return out\n    ",
      "execution_count": 15,
      "outputs": []
    },
    {
      "metadata": {
        "_uuid": "b2a2b143d8eee3045019748a02b4ffacb50828f3"
      },
      "cell_type": "markdown",
      "source": "Create the Deep NN object cnn  = CNN_sign( ) and we are using Cross rntropy loss in this case . Optimizer  we are choosing stochastic gradient descent (SGD) with a learning rate of 0.02 but Adam would have also been good perfoming here."
    },
    {
      "metadata": {
        "scrolled": true,
        "_uuid": "5e4fcfb49eb87db66e8b5d1bd947a9a32bfd2155",
        "_cell_guid": "8d050f8f-b2bc-4bcc-aee5-648a0a5985a7",
        "trusted": true
      },
      "cell_type": "code",
      "source": "cnn = CNN_sign()\nloss_fn = nn.CrossEntropyLoss()\ncnn_param = cnn.parameters()\noptimizer = torch.optim.SGD(list(cnn_param), lr=0.02)\nprint(cnn)",
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": "CNN_sign (\n  (layer1): Sequential (\n    (0): Conv2d(3, 10, kernel_size=(5, 5), stride=(1, 1))\n    (1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True)\n    (2): ReLU ()\n    (3): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))\n    (4): Conv2d(10, 10, kernel_size=(5, 5), stride=(1, 1))\n    (5): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True)\n    (6): ReLU ()\n    (7): MaxPool2d (size=(2, 2), stride=(2, 2), dilation=(1, 1))\n  )\n  (fc1): Linear (1690 -> 100)\n  (fc2): BatchNorm1d(100, eps=1e-05, momentum=0.1, affine=True)\n  (fc3): Linear (100 -> 6)\n)\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "2f7f884ec092eeab7641b5db838ca55f027f3a41"
      },
      "cell_type": "markdown",
      "source": "We are just running this deep NN for only 10 epochs as it takes quite a few minutes to get executed .  "
    },
    {
      "metadata": {
        "_uuid": "57c1a284aaf0ead88b15e3423eb6c5b4fd47d71a",
        "_cell_guid": "374c2c56-65ad-4e24-a327-a235a16431ce",
        "trusted": true
      },
      "cell_type": "code",
      "source": "num_epochs = 10\nfor epoch in range(num_epochs):\n    for i,(image,label) in enumerate(train_loader):\n        image = Variable(image) \n        label = Variable(label)\n        \n        output = cnn(image) # forward propagation to produce output\n        loss = loss_fn(output,label) # find the loss\n        \n        loss.backward() # find the gradients using back propagation\n        optimizer.step() # updating the parameters\n        \n        #if (i+1) % 20 == 0:\n        #print ('Epoch [%d/%d], Iter [%d/%d] Loss: %.4f' %(epoch+1, 10, i+1, len(train_data)//64 ,  loss.data[0]))\n    print ('Epoch [%d/%d], Loss: %.4f' %(epoch+1, 10,  loss.data[0]))",
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": "Epoch [1/10], Loss: 1.3127\nEpoch [2/10], Loss: 0.5887\nEpoch [3/10], Loss: 0.6348\nEpoch [4/10], Loss: 0.6245\nEpoch [5/10], Loss: 0.3420\nEpoch [6/10], Loss: 0.7494\nEpoch [7/10], Loss: 0.1952\nEpoch [8/10], Loss: 0.1748\nEpoch [9/10], Loss: 0.2643\nEpoch [10/10], Loss: 0.2444\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "34e0984c27323bd0a8593e09db86de3490e60569"
      },
      "cell_type": "markdown",
      "source": "Now first we test the model on our training data ."
    },
    {
      "metadata": {
        "_uuid": "bc5eff51045b87f93bdc3096a8608b5d48c77fac",
        "_cell_guid": "311824a3-51ac-4562-b6f0-7a4f03cebe33",
        "trusted": true
      },
      "cell_type": "code",
      "source": "pred = cnn(Variable(torch.Tensor((train_x.reshape(1080,3,64,64)).astype(float))))\npred_np = pred.data.numpy()\npred_label = np.argmax(pred_np,axis=1)\npred_label.shape",
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 18,
          "data": {
            "text/plain": "(1080,)"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "b57acbbe8450281883a40e44ed8f1a51ae22a959"
      },
      "cell_type": "markdown",
      "source": "This is our train accuracy 92.6 %"
    },
    {
      "metadata": {
        "_uuid": "89f978d1b8476c5341843581de0197ef475504ed",
        "_cell_guid": "bc439e5a-9d5a-49b3-8b77-694b8a5ea621",
        "trusted": true
      },
      "cell_type": "code",
      "source": "target = np.squeeze(train_y)\nnp.sum(pred_label==target)/1080",
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 19,
          "data": {
            "text/plain": "0.92592592592592593"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "7332b2272af11a465c7e8356ac04451f13b34ecb"
      },
      "cell_type": "markdown",
      "source": "Classification report on training data "
    },
    {
      "metadata": {
        "_uuid": "d85b35d6e78f39f1a7b75e8f84107600ba9afa4a",
        "_cell_guid": "fa09893a-d13f-4cb7-81fe-cab9a98ef19e",
        "trusted": true
      },
      "cell_type": "code",
      "source": "from sklearn.metrics import classification_report\nprint(classification_report(target,pred_label))",
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": "             precision    recall  f1-score   support\n\n          0       0.96      0.99      0.98       180\n          1       0.99      0.83      0.91       180\n          2       0.89      0.94      0.91       180\n          3       0.92      0.98      0.95       180\n          4       0.88      0.95      0.91       180\n          5       0.93      0.86      0.89       180\n\navg / total       0.93      0.93      0.93      1080\n\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "611e9b9fe3c7deafd0aac11dfd6e43414f23acca"
      },
      "cell_type": "markdown",
      "source": "Testing on the test data , which gives almost 86 % accuracy ."
    },
    {
      "metadata": {
        "_uuid": "512d4cf2b77ca4c2f4b7bc25c8b56ad13c28adae",
        "_cell_guid": "55bde02a-2af5-4226-bed7-139871d97b78",
        "trusted": true
      },
      "cell_type": "code",
      "source": "pred = cnn(Variable(torch.Tensor((test_x.reshape(120,3,64,64)).astype(float))))\npred_np = pred.data.numpy()\npred_label = np.argmax(pred_np,axis=1)\npred_label.shape\ntarget = np.squeeze(test_y)\nnp.sum(pred_label==target)/120",
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 22,
          "data": {
            "text/plain": "0.85833333333333328"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "_uuid": "d701e74b40a9f729efdee4bb51b0ff5e6432ce92"
      },
      "cell_type": "markdown",
      "source": "Classification report on test data . "
    },
    {
      "metadata": {
        "trusted": true,
        "_uuid": "aac5ed063a1f27418710315fec3ab3184a7efea1"
      },
      "cell_type": "code",
      "source": "print(classification_report(target,pred_label))",
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": "             precision    recall  f1-score   support\n\n          0       0.87      1.00      0.93        20\n          1       1.00      0.80      0.89        20\n          2       0.85      0.85      0.85        20\n          3       0.82      0.90      0.86        20\n          4       0.80      0.80      0.80        20\n          5       0.84      0.80      0.82        20\n\navg / total       0.86      0.86      0.86       120\n\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "collapsed": true,
        "_uuid": "ffba2675001d848040024281abbdfa0b4c844061"
      },
      "cell_type": "code",
      "source": "",
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "name": "python",
      "version": "3.6.3",
      "mimetype": "text/x-python",
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "pygments_lexer": "ipython3",
      "nbconvert_exporter": "python",
      "file_extension": ".py"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 1
}