FocalLoss.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "06aabb63",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T16:48:01.248865Z",
     "start_time": "2021-09-20T16:47:59.029865Z"
    },
    "hide_input": false
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch import nn\n",
    "import pretrainedmodels\n",
    "\n",
    "from utils.train import train\n",
    "from utils.evaluation import evaluation\n",
    "from utils.loader import get_loader, get_image_list\n",
    "from utils.losses import CostSensitiveRegularizedLoss\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\", category=UserWarning)\n",
    "warnings.filterwarnings(\"ignore\", category=DeprecationWarning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "cb64b0db",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T16:48:01.264865Z",
     "start_time": "2021-09-20T16:48:01.249867Z"
    }
   },
   "outputs": [],
   "source": [
    "# hyper params\n",
    "num_of_class = 102\n",
    "learning_rate = 1e-03\n",
    "batch_size = 128\n",
    "input_size = 224\n",
    "device = torch.device(\"cuda\")\n",
    "test_size = 0.2\n",
    "train_size = 0.7\n",
    "val_size = 0.1\n",
    "directoryAgeDB = 'AgeDB/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "6d920459",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T16:48:01.280866Z",
     "start_time": "2021-09-20T16:48:01.265865Z"
    },
    "code_folding": [],
    "hide_input": false
   },
   "outputs": [],
   "source": [
    "def findAge(image_list):\n",
    "    listofzeros = [0] * 102\n",
    "    same_age = [i for i in range(0, 102)]\n",
    "    age_number = dict(zip(same_age, listofzeros))\n",
    "    sum = 0\n",
    "\n",
    "    for i in range(len(image_list)):\n",
    "\n",
    "        if image_list[i]['age'] in age_number:\n",
    "            age_number[image_list[i]['age']] += 1\n",
    "            sum += 1\n",
    "\n",
    "    print(sum)\n",
    "    return dict(age_number)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "f8ae2b44",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T16:49:38.208870Z",
     "start_time": "2021-09-20T16:48:01.281865Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c2d855dd55ec4e27a9f739030ce36d14",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/11541 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Set Length:  34623\n",
      "Test Set Length:  3298\n",
      "Validation Set Length:  1649\n",
      "Total:  39570\n"
     ]
    }
   ],
   "source": [
    "train_image_list, test_image_list, validate_image_list = get_image_list(\n",
    "    train_size=train_size, \n",
    "    test_size=test_size, \n",
    "    val_size=val_size, \n",
    "    directory=directoryAgeDB,\n",
    ")\n",
    "\n",
    "train_loader, test_loader, validation_loader = get_loader(\n",
    "    input_size=input_size,\n",
    "    train_image_list=train_image_list,\n",
    "    test_image_list=test_image_list, \n",
    "    validate_image_list=validate_image_list, \n",
    "    batch_size=batch_size,\n",
    "    train_augment=3,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "5641b0b5",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T16:49:41.188978Z",
     "start_time": "2021-09-20T16:49:38.213866Z"
    },
    "code_folding": []
   },
   "outputs": [],
   "source": [
    "resnet1 = pretrainedmodels.resnet18(pretrained='imagenet')\n",
    "resnet1.last_linear = nn.Linear(\n",
    "    in_features=resnet1.last_linear.in_features, \n",
    "    out_features=num_of_class, \n",
    "    bias=False\n",
    ")\n",
    "resnet2 = pretrainedmodels.resnet18(pretrained='imagenet')\n",
    "resnet2.last_linear = nn.Linear(\n",
    "    in_features=resnet2.last_linear.in_features, \n",
    "    out_features=num_of_class, \n",
    "    bias=False\n",
    ")\n",
    "resnet1 = resnet1.to(device)\n",
    "resnet2 = resnet2.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "3ca9b2a2",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T16:49:41.204980Z",
     "start_time": "2021-09-20T16:49:41.191980Z"
    }
   },
   "outputs": [],
   "source": [
    "criteria = CostSensitiveRegularizedLoss(n_classes=num_of_class, base_loss='focal_loss', reduction='mean') \n",
    "optimizer = torch.optim.SGD(resnet1.parameters(), lr=learning_rate, momentum=0.9, nesterov=True, weight_decay=1e-1)\n",
    "model_name = 'focalloss-aug.pth'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "6ed4bb3e",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T17:00:55.074039Z",
     "start_time": "2021-09-20T16:49:41.209980Z"
    },
    "hide_input": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1/10, Train Loss: 2.5497877395460966,  Validation Loss: 2.3605024814605713\n",
      "Validation Loss Decreased(inf ---> 2.3605024814605713)\n",
      "Epoch: 2/10, Train Loss: 2.1918997051970983,  Validation Loss: 2.1458323001861572\n",
      "Validation Loss Decreased(2.3605024814605713 ---> 2.1458323001861572)\n",
      "Epoch: 3/10, Train Loss: 2.0429964017164224,  Validation Loss: 2.068101708705609\n",
      "Validation Loss Decreased(2.1458323001861572 ---> 2.068101708705609)\n",
      "Epoch: 4/10, Train Loss: 1.9810232434325552,  Validation Loss: 2.0599818779872012\n",
      "Validation Loss Decreased(2.068101708705609 ---> 2.0599818779872012)\n",
      "Epoch: 5/10, Train Loss: 1.9675067557619947,  Validation Loss: 2.0964393524023204\n",
      "Epoch: 6/10, Train Loss: 1.9959682067821827,  Validation Loss: 2.165790374462421\n",
      "Epoch: 7/10, Train Loss: 2.0598285180616203,  Validation Loss: 2.255103881542499\n",
      "Epoch: 8/10, Train Loss: 2.1588039811686834,  Validation Loss: 2.347843041786781\n",
      "Epoch: 9/10, Train Loss: 2.253788910228828,  Validation Loss: 2.40657272705665\n",
      "Epoch: 10/10, Train Loss: 2.3304923654042486,  Validation Loss: 2.474382969049307\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train(resnet1, optimizer, criteria, train_loader, validation_loader, num_of_epoch=10, model_name=model_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "10b735ef",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T17:01:23.097948Z",
     "start_time": "2021-09-20T17:00:55.076043Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Absolute Error: 7.3978166580200195\n",
      "Minimum: 0.0, Maximum: 48.0, Median: 6.0\n",
      "Mean Absolute Error: 6.835201644897461\n",
      "Minimum: 0.0, Maximum: 49.0, Median: 5.0\n"
     ]
    }
   ],
   "source": [
    "resnet1.load_state_dict(torch.load(model_name))\n",
    "loss_per_class = evaluation(resnet1, test_loader)\n",
    "_ = evaluation(resnet1, train_loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "6452b440",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T17:01:23.158948Z",
     "start_time": "2021-09-20T17:01:23.098949Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3298\n",
      "Age: Test Sample: Loss\n",
      "1 : 1 : 34.00\n",
      "2  No Sample\n",
      "3 : 2 : 28.50\n",
      "4 : 1 : 31.00\n",
      "5 : 1 : 23.00\n",
      "6 : 2 : 22.00\n",
      "7 : 2 : 22.00\n",
      "8 : 2 : 23.50\n",
      "9 : 1 : 19.00\n",
      "10 : 3 : 22.67\n",
      "11 : 4 : 18.75\n",
      "12 : 3 : 17.33\n",
      "13 : 4 : 18.00\n",
      "14 : 5 : 16.20\n",
      "15 : 9 : 13.44\n",
      "16 : 10 : 12.80\n",
      "17 : 16 : 11.81\n",
      "18 : 24 : 11.62\n",
      "19 : 22 : 10.00\n",
      "20 : 32 : 10.12\n",
      "21 : 33 : 8.06\n",
      "22 : 37 : 8.65\n",
      "23 : 42 : 7.26\n",
      "24 : 52 : 6.92\n",
      "25 : 57 : 4.93\n",
      "26 : 55 : 4.07\n",
      "27 : 61 : 4.77\n",
      "28 : 69 : 3.75\n",
      "29 : 62 : 3.90\n",
      "30 : 77 : 3.23\n",
      "31 : 63 : 5.13\n",
      "32 : 75 : 4.91\n",
      "33 : 67 : 4.69\n",
      "34 : 72 : 5.44\n",
      "35 : 82 : 4.73\n",
      "36 : 69 : 5.12\n",
      "37 : 74 : 6.27\n",
      "38 : 76 : 5.83\n",
      "39 : 64 : 6.22\n",
      "40 : 77 : 6.34\n",
      "41 : 64 : 6.08\n",
      "42 : 71 : 7.20\n",
      "43 : 65 : 7.65\n",
      "44 : 56 : 8.07\n",
      "45 : 68 : 7.01\n",
      "46 : 57 : 7.91\n",
      "47 : 58 : 7.07\n",
      "48 : 62 : 8.08\n",
      "49 : 53 : 8.83\n",
      "50 : 63 : 8.51\n",
      "51 : 57 : 8.00\n",
      "52 : 62 : 9.15\n",
      "53 : 50 : 7.50\n",
      "54 : 51 : 8.39\n",
      "55 : 59 : 9.88\n",
      "56 : 45 : 7.73\n",
      "57 : 49 : 9.41\n",
      "58 : 49 : 7.96\n",
      "59 : 43 : 8.28\n",
      "60 : 49 : 7.53\n",
      "61 : 41 : 8.32\n",
      "62 : 48 : 7.77\n",
      "63 : 44 : 6.98\n",
      "64 : 39 : 8.21\n",
      "65 : 46 : 5.89\n",
      "66 : 35 : 6.60\n",
      "67 : 35 : 7.80\n",
      "68 : 43 : 7.42\n",
      "69 : 30 : 6.90\n",
      "70 : 43 : 4.98\n",
      "71 : 29 : 4.14\n",
      "72 : 35 : 6.37\n",
      "73 : 31 : 5.42\n",
      "74 : 28 : 10.36\n",
      "75 : 32 : 9.78\n",
      "76 : 22 : 9.64\n",
      "77 : 21 : 8.57\n",
      "78 : 21 : 11.86\n",
      "79 : 15 : 9.33\n",
      "80 : 18 : 10.83\n",
      "81 : 15 : 12.33\n",
      "82 : 14 : 14.00\n",
      "83 : 11 : 13.18\n",
      "84 : 9 : 16.56\n",
      "85 : 12 : 17.50\n",
      "86 : 7 : 16.71\n",
      "87 : 6 : 19.50\n",
      "88 : 4 : 18.00\n",
      "89 : 4 : 19.00\n",
      "90 : 4 : 21.25\n",
      "91 : 3 : 21.00\n",
      "92 : 2 : 22.00\n",
      "93 : 3 : 23.00\n",
      "94 : 1 : 24.00\n",
      "95 : 2 : 25.00\n",
      "96 : 1 : 26.00\n",
      "97 : 1 : 27.00\n",
      "98 : 1 : 28.00\n",
      "99 : 1 : 29.00\n",
      "100 : 1 : 38.00\n",
      "101 : 1 : 31.00\n"
     ]
    }
   ],
   "source": [
    "sample_per_class = findAge(test_image_list)\n",
    "print('Age: Test Sample: Loss')\n",
    "for i in range(1, 102):\n",
    "    if sample_per_class[i] == 0:\n",
    "        print(i,' No Sample')\n",
    "    else:\n",
    "        print(i,':', sample_per_class[i],':', format(loss_per_class[i].item()/sample_per_class[i],'.2f'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "052a2ce7",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T17:11:58.512250Z",
     "start_time": "2021-09-20T17:07:34.092698Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "94d7ddb332dd47848f2c2ff8e1141380",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/11541 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Set Length:  11541\n",
      "Test Set Length:  3298\n",
      "Validation Set Length:  1649\n",
      "Total:  16488\n",
      "Epoch: 1/10, Train Loss: 1.9113213085866236,  Validation Loss: 2.064310440650353\n",
      "Validation Loss Decreased(inf ---> 2.064310440650353)\n",
      "Epoch: 2/10, Train Loss: 1.8834970442803352,  Validation Loss: 2.068317248271062\n",
      "Epoch: 3/10, Train Loss: 1.860955834388733,  Validation Loss: 2.0755570851839504\n",
      "Epoch: 4/10, Train Loss: 1.8404606397335346,  Validation Loss: 2.089746356010437\n",
      "Epoch: 5/10, Train Loss: 1.8397495602513407,  Validation Loss: 2.106857703282283\n",
      "Epoch: 6/10, Train Loss: 1.8406613305374817,  Validation Loss: 2.188012434886052\n",
      "Epoch: 7/10, Train Loss: 1.8590041464501685,  Validation Loss: 2.1494343280792236\n",
      "Epoch: 8/10, Train Loss: 1.8779431188499534,  Validation Loss: 2.1969324075258694\n",
      "Epoch: 9/10, Train Loss: 1.9133659051014826,  Validation Loss: 2.2028550184690037\n",
      "Epoch: 10/10, Train Loss: 1.9406616072078327,  Validation Loss: 2.220918765434852\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_loader, test_loader, validation_loader = get_loader(\n",
    "    input_size=input_size,\n",
    "    train_image_list=train_image_list,\n",
    "    test_image_list=test_image_list, \n",
    "    validate_image_list=validate_image_list, \n",
    "    batch_size=batch_size,\n",
    "    train_augment=1,\n",
    ")\n",
    "criteria = CostSensitiveRegularizedLoss(n_classes=num_of_class, base_loss='focal_loss', reduction='mean') \n",
    "optimizer = torch.optim.SGD(resnet2.parameters(), lr=learning_rate, momentum=0.9, nesterov=True, weight_decay=1e-1)\n",
    "model_name = 'focalloss.pth'\n",
    "train(resnet2, optimizer, criteria, train_loader, validation_loader, num_of_epoch=10, model_name=model_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "86013069",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T17:15:23.569976Z",
     "start_time": "2021-09-20T17:15:03.384975Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Absolute Error: 7.603395938873291\n",
      "Minimum: 0.0, Maximum: 50.0, Median: 6.0\n",
      "Mean Absolute Error: 5.7506279945373535\n",
      "Minimum: 0.0, Maximum: 45.0, Median: 5.0\n"
     ]
    }
   ],
   "source": [
    "resnet2.load_state_dict(torch.load(model_name))\n",
    "loss_per_class = evaluation(resnet2, test_loader)\n",
    "_ = evaluation(resnet2, train_loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "33c7b360",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-09-20T17:15:27.920110Z",
     "start_time": "2021-09-20T17:15:27.795113Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3298\n",
      "Age: Test Sample: Loss\n",
      "1 : 1 : 39.00\n",
      "2  No Sample\n",
      "3 : 2 : 35.50\n",
      "4 : 1 : 31.00\n",
      "5 : 1 : 23.00\n",
      "6 : 2 : 22.00\n",
      "7 : 2 : 22.00\n",
      "8 : 2 : 23.50\n",
      "9 : 1 : 21.00\n",
      "10 : 3 : 26.00\n",
      "11 : 4 : 21.00\n",
      "12 : 3 : 18.67\n",
      "13 : 4 : 19.75\n",
      "14 : 5 : 17.00\n",
      "15 : 9 : 14.67\n",
      "16 : 10 : 13.70\n",
      "17 : 16 : 12.00\n",
      "18 : 24 : 12.33\n",
      "19 : 22 : 10.36\n",
      "20 : 32 : 10.81\n",
      "21 : 33 : 8.73\n",
      "22 : 37 : 8.59\n",
      "23 : 42 : 7.38\n",
      "24 : 52 : 6.87\n",
      "25 : 57 : 5.25\n",
      "26 : 55 : 5.58\n",
      "27 : 61 : 4.95\n",
      "28 : 69 : 4.28\n",
      "29 : 62 : 4.27\n",
      "30 : 77 : 3.06\n",
      "31 : 63 : 4.60\n",
      "32 : 75 : 4.92\n",
      "33 : 67 : 4.30\n",
      "34 : 72 : 5.24\n",
      "35 : 82 : 5.50\n",
      "36 : 69 : 5.13\n",
      "37 : 74 : 6.05\n",
      "38 : 76 : 5.83\n",
      "39 : 64 : 6.06\n",
      "40 : 77 : 5.56\n",
      "41 : 64 : 6.12\n",
      "42 : 71 : 6.41\n",
      "43 : 65 : 7.42\n",
      "44 : 56 : 7.75\n",
      "45 : 68 : 7.31\n",
      "46 : 57 : 8.23\n",
      "47 : 58 : 7.14\n",
      "48 : 62 : 9.27\n",
      "49 : 53 : 9.60\n",
      "50 : 63 : 8.51\n",
      "51 : 57 : 8.47\n",
      "52 : 62 : 9.23\n",
      "53 : 50 : 7.68\n",
      "54 : 51 : 7.47\n",
      "55 : 59 : 10.81\n",
      "56 : 45 : 8.00\n",
      "57 : 49 : 9.27\n",
      "58 : 49 : 8.12\n",
      "59 : 43 : 8.81\n",
      "60 : 49 : 7.92\n",
      "61 : 41 : 7.80\n",
      "62 : 48 : 7.46\n",
      "63 : 44 : 7.30\n",
      "64 : 39 : 8.26\n",
      "65 : 46 : 5.80\n",
      "66 : 35 : 8.91\n",
      "67 : 35 : 7.91\n",
      "68 : 43 : 7.72\n",
      "69 : 30 : 7.50\n",
      "70 : 43 : 7.58\n",
      "71 : 29 : 4.48\n",
      "72 : 35 : 6.03\n",
      "73 : 31 : 5.74\n",
      "74 : 28 : 11.00\n",
      "75 : 32 : 10.69\n",
      "76 : 22 : 9.32\n",
      "77 : 21 : 8.19\n",
      "78 : 21 : 12.52\n",
      "79 : 15 : 10.27\n",
      "80 : 18 : 10.78\n",
      "81 : 15 : 13.00\n",
      "82 : 14 : 14.14\n",
      "83 : 11 : 13.45\n",
      "84 : 9 : 16.22\n",
      "85 : 12 : 18.50\n",
      "86 : 7 : 17.43\n",
      "87 : 6 : 19.50\n",
      "88 : 4 : 18.00\n",
      "89 : 4 : 19.00\n",
      "90 : 4 : 20.00\n",
      "91 : 3 : 21.67\n",
      "92 : 2 : 22.00\n",
      "93 : 3 : 23.67\n",
      "94 : 1 : 29.00\n",
      "95 : 2 : 25.00\n",
      "96 : 1 : 26.00\n",
      "97 : 1 : 27.00\n",
      "98 : 1 : 28.00\n",
      "99 : 1 : 29.00\n",
      "100 : 1 : 35.00\n",
      "101 : 1 : 33.00\n"
     ]
    }
   ],
   "source": [
    "sample_per_class = findAge(test_image_list)\n",
    "print('Age: Test Sample: Loss')\n",
    "for i in range(1, 102):\n",
    "    if sample_per_class[i] == 0:\n",
    "        print(i,' No Sample')\n",
    "    else:\n",
    "        print(i,':', sample_per_class[i],':', format(loss_per_class[i].item()/sample_per_class[i],'.2f'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cf0b1434",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}