example_data/model_generation/LeafSnap/generate_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "random-strain",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import tqdm\n",
    "\n",
    "import wandb\n",
    "import torch\n",
    "from torch import nn, optim\n",
    "import torch.nn.functional as F\n",
    "from torchvision import datasets, transforms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "1640730a-090d-46e6-a4c1-bbeeb03f6826",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mode = 'disabled'\n",
    "# mode = 'online'\n",
    "\n",
    "wandb.init(project='leafsnap', entity='dianna-ai', mode=mode)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "bb15482a-1c87-4d05-ab9a-81a507e49512",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Note: these hyperparameter values were updated to be equal to the optimal ones chosen by W&B,\n",
    "# the rest of this notebook still uses old values, that were just and example,\n",
    "config = wandb.config\n",
    "config.img_size = 128\n",
    "config.batch_size = 128\n",
    "config.epochs = 10\n",
    "config.learning_rate = 0.000359\n",
    "config.weight_decay = 0.00181\n",
    "config.kernel_size = 3  # must be an odd number to keep the image dimensions the same after convolution\n",
    "config.pooling = 2\n",
    "config.channels1 = 62\n",
    "config.channels2 = 55\n",
    "config.channels3 = 40\n",
    "config.channels4 = 42\n",
    "config.dropout = .197"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "473133b6-0704-4446-99ec-d3fdfe6cc572",
   "metadata": {},
   "outputs": [],
   "source": [
    "if torch.cuda.is_available():\n",
    "    device = torch.device('cuda')\n",
    "else:\n",
    "    device = torch.device('cpu')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "quantitative-accreditation",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>file_id</th>\n",
       "      <th>image_path</th>\n",
       "      <th>species</th>\n",
       "      <th>source</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>77937</td>\n",
       "      <td>dataset/images/lab/Auto_cropped/diospyros_virg...</td>\n",
       "      <td>Diospyros virginiana</td>\n",
       "      <td>lab</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>77938</td>\n",
       "      <td>dataset/images/lab/Auto_cropped/diospyros_virg...</td>\n",
       "      <td>Diospyros virginiana</td>\n",
       "      <td>lab</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>77939</td>\n",
       "      <td>dataset/images/lab/Auto_cropped/diospyros_virg...</td>\n",
       "      <td>Diospyros virginiana</td>\n",
       "      <td>lab</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>77940</td>\n",
       "      <td>dataset/images/lab/Auto_cropped/diospyros_virg...</td>\n",
       "      <td>Diospyros virginiana</td>\n",
       "      <td>lab</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>77941</td>\n",
       "      <td>dataset/images/lab/Auto_cropped/diospyros_virg...</td>\n",
       "      <td>Diospyros virginiana</td>\n",
       "      <td>lab</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   file_id                                         image_path  \\\n",
       "0    77937  dataset/images/lab/Auto_cropped/diospyros_virg...   \n",
       "1    77938  dataset/images/lab/Auto_cropped/diospyros_virg...   \n",
       "2    77939  dataset/images/lab/Auto_cropped/diospyros_virg...   \n",
       "3    77940  dataset/images/lab/Auto_cropped/diospyros_virg...   \n",
       "4    77941  dataset/images/lab/Auto_cropped/diospyros_virg...   \n",
       "\n",
       "                species source  \n",
       "0  Diospyros virginiana    lab  \n",
       "1  Diospyros virginiana    lab  \n",
       "2  Diospyros virginiana    lab  \n",
       "3  Diospyros virginiana    lab  \n",
       "4  Diospyros virginiana    lab  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset_root = os.path.expanduser('~/nlesc/DIANNA/datasets/leafsnap/leafsnap-dataset-30subset')\n",
    "\n",
    "metadata = pd.read_csv(os.path.join(dataset_root, 'leafsnap-dataset-30subset-images.txt'), delimiter='\\t')\n",
    "metadata.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "50b37f89-dd63-4bc3-bfb8-4ccfe30df53c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of classes: 30\n",
      "Total number of samples: 7395\n",
      "Number of training samples: 5917\n",
      "Number of validation samples: 739\n",
      "Number of test samples: 739\n"
     ]
    }
   ],
   "source": [
    "def valid_file(fname):\n",
    "    # dataset consists of .jpg images\n",
    "    # converted to .bmp for faster loading\n",
    "    return fname.endswith('.bmp')\n",
    "\n",
    "transform = transforms.Compose([transforms.Resize(config.img_size),\n",
    "                                transforms.CenterCrop(config.img_size),\n",
    "                                transforms.ToTensor()])\n",
    "\n",
    "train_data = datasets.ImageFolder(os.path.join(dataset_root, 'dataset/split/train'), transform=transform, is_valid_file=valid_file)\n",
    "val_data = datasets.ImageFolder(os.path.join(dataset_root, 'dataset/split/validation'), transform=transform, is_valid_file=valid_file)\n",
    "test_data = datasets.ImageFolder(os.path.join(dataset_root, 'dataset/split/test'), transform=transform, is_valid_file=valid_file)\n",
    "\n",
    "num_classes = len(train_data.classes)\n",
    "num_samples = len(train_data) + len(val_data) + len(test_data)\n",
    "\n",
    "print(f'Number of classes: {num_classes}')\n",
    "print(f'Total number of samples: {num_samples}')\n",
    "print(f'Number of training samples: {len(train_data)}')\n",
    "print(f'Number of validation samples: {len(val_data)}')\n",
    "print(f'Number of test samples: {len(test_data)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "defined-campus",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'ulmus_pumila')"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "workers = min(12, os.cpu_count() - 1)\n",
    "\n",
    "train_dataloader = torch.utils.data.DataLoader(train_data, batch_size=config.batch_size, shuffle=True, num_workers=workers, pin_memory=True)\n",
    "val_dataloader = torch.utils.data.DataLoader(val_data, batch_size=config.batch_size, shuffle=True, num_workers=workers, pin_memory=True)\n",
    "test_dataloader = torch.utils.data.DataLoader(test_data, batch_size=config.batch_size, shuffle=True, num_workers=workers, pin_memory=True)\n",
    "\n",
    "images, labels = next(iter(train_dataloader))\n",
    "im = np.array(images[1].permute(1,2,0))\n",
    "plt.imshow(im)\n",
    "plt.title(train_data.classes[labels[0]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "coupled-aruba",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of trainable parameters: 210666\n"
     ]
    }
   ],
   "source": [
    "class Network(nn.Module):\n",
    "    def __init__(self, img_size, num_classes, kernel_size, pooling_size, num_channels, dropout):\n",
    "        super().__init__()\n",
    "        \n",
    "        self.num = len(num_channels)\n",
    "        \n",
    "        self.conv_layers = nn.ModuleList()\n",
    "        self.bn_layers = nn.ModuleList()\n",
    "        self.pool_layers = nn.ModuleList()\n",
    "        \n",
    "        for i, out_channels in enumerate(num_channels):\n",
    "            if i == 0:\n",
    "                in_channels = 3\n",
    "            else:\n",
    "                in_channels = num_channels[i-1]\n",
    "            padding = int((kernel_size - 1) / 2)\n",
    "            self.conv_layers.append(nn.Conv2d(in_channels=in_channels, out_channels=out_channels, kernel_size=kernel_size, stride=1, padding=padding))\n",
    "            self.bn_layers.append(nn.BatchNorm2d(out_channels))\n",
    "            # no pooling layer for the last step\n",
    "            if i < self.num - 1:\n",
    "                self.pool_layers.append(nn.MaxPool2d(pooling_size, pooling_size))\n",
    "        \n",
    "        # image size has been reduced by a factor pooling_size as many times as there are pooling layres\n",
    "        reduced_img_size = int(img_size / pooling_size ** len(self.pool_layers))\n",
    "        self.final_layer_size = num_channels[-1] * reduced_img_size ** 2\n",
    "\n",
    "        self.drop = nn.Dropout(dropout)\n",
    "\n",
    "        self.fc1 = nn.Linear(self.final_layer_size, num_classes)        \n",
    "\n",
    "    def forward(self, data):\n",
    "        for i in range(self.num):\n",
    "            conv = self.conv_layers[i]\n",
    "            bn = self.bn_layers[i]\n",
    "            data = F.relu(bn(conv(data)))\n",
    "            # apply pooling except in the last step\n",
    "            if i < self.num - 1:\n",
    "                pool = self.pool_layers[i]\n",
    "                data = pool(data)\n",
    "\n",
    "        data = data.view(-1, self.final_layer_size)\n",
    "        data = self.fc1(self.drop(data))\n",
    "        return data\n",
    "    \n",
    "# Instantiate a neural network model \n",
    "channels = [config.channels1, config.channels2, config.channels3, config.channels4]\n",
    "model = Network(config.img_size, num_classes, config.kernel_size, config.pooling, channels, config.dropout).to(device)\n",
    "\n",
    "print(f'Number of trainable parameters: {sum([p.numel() for p in model.parameters()])}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "7cb63cba-b61d-407c-8a5a-a583490bbce3",
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam(model.parameters(), lr=config.learning_rate, weight_decay=config.weight_decay)\n",
    "loss_func = nn.CrossEntropyLoss().to(device)\n",
    "\n",
    "\n",
    "def accuracy(model_output, y_true):\n",
    "    _, y_pred = torch.max(model_output, dim=1)\n",
    "    return (y_pred == y_true).sum() / len(y_pred)\n",
    "\n",
    "\n",
    "def train(model, train_data, optimizer, loss_func):\n",
    "    epoch_loss = 0\n",
    "    epoch_acc = 0\n",
    "    model.train()\n",
    "\n",
    "    for batch in tqdm.notebook.tqdm(train_data):\n",
    "        images, labels = batch\n",
    "        images = images.to(device)\n",
    "        labels = labels.to(device)\n",
    "        \n",
    "        optimizer.zero_grad()\n",
    "        predictions = model(images)\n",
    "       \n",
    "        loss = loss_func(predictions, labels)\n",
    "        acc = accuracy(predictions, labels)\n",
    "        \n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        \n",
    "        epoch_loss += loss.item()\n",
    "        epoch_acc += acc.item()\n",
    "        \n",
    "        wandb.log({'train_loss': loss.item(), 'train_acc': acc.item()})\n",
    "\n",
    "    epoch_loss /= len(train_data)\n",
    "    epoch_acc /= len(train_data)\n",
    "    \n",
    "    return epoch_loss, epoch_acc\n",
    "\n",
    "\n",
    "@torch.no_grad()\n",
    "def evaluate(model, data, loss_func, train_mode=False):\n",
    "    avg_loss = 0\n",
    "    avg_acc = 0\n",
    "    if train_mode:\n",
    "        model.train()\n",
    "    else:\n",
    "        model.eval()\n",
    "        \n",
    "    for batch in data:\n",
    "        images, labels = batch\n",
    "        images = images.to(device)\n",
    "        labels = labels.to(device)\n",
    "        \n",
    "        predictions = model(images)\n",
    "       \n",
    "        loss = loss_func(predictions, labels)\n",
    "        acc = accuracy(predictions, labels)\n",
    "        \n",
    "        avg_loss += loss.item()\n",
    "        avg_acc += acc.item()\n",
    "\n",
    "    avg_loss /= len(data)\n",
    "    avg_acc /= len(data)\n",
    "    \n",
    "    return avg_loss, avg_acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "d1f6d8b4-ce21-4bf6-9529-a3c5d8aa8208",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0c6615eb01554496a53c2044524f67a6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 2.22 | train acc: 0.35\n",
      "val   loss: 1.96 | val   acc: 0.42\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5b902bbd29754429ae0a74d788fd5d80",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 1.42 | train acc: 0.55\n",
      "val   loss: 2.25 | val   acc: 0.42\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d3d50d631b5b4e158e8f7b145358b962",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 1.15 | train acc: 0.64\n",
      "val   loss: 1.27 | val   acc: 0.59\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "97c907417fab40b394b21f81578a8d39",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 0.92 | train acc: 0.71\n",
      "val   loss: 3.00 | val   acc: 0.41\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1410f260ed0a473fbaf0f16a3c20e8fc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 0.78 | train acc: 0.75\n",
      "val   loss: 1.24 | val   acc: 0.63\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "66e320cd8d884d5d8606cbb52fc0e037",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 0.67 | train acc: 0.78\n",
      "val   loss: 3.47 | val   acc: 0.39\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "11d723b8a72e43e09722b7992860e67e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 0.57 | train acc: 0.81\n",
      "val   loss: 1.23 | val   acc: 0.62\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "668087df65d448748ccb269770ad3783",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 0.50 | train acc: 0.83\n",
      "val   loss: 0.91 | val   acc: 0.72\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e2b241d1d6a0418c9bae290a65ea1320",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 0.43 | train acc: 0.86\n",
      "val   loss: 1.73 | val   acc: 0.61\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6128e28017894ad5a133ae5dfe6089ac",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/185 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 0.41 | train acc: 0.86\n",
      "val   loss: 0.79 | val   acc: 0.76\n",
      "best val loss: 0.79 | best val acc: 0.76\n"
     ]
    }
   ],
   "source": [
    "best_val_loss = np.inf\n",
    "best_val_acc = None\n",
    "\n",
    "for epoch in range(config.epochs):\n",
    "    # run training loop\n",
    "    train_loss, train_acc = train(model, train_dataloader, optimizer, loss_func)\n",
    "    # evalute on validation data, with same model settings as during training (i.e. including dropout)\n",
    "    val_loss, val_acc = evaluate(model, val_dataloader, loss_func)\n",
    "\n",
    "    wandb.log({'train_epoch_loss': train_loss, 'train_epoch_acc': train_acc})\n",
    "    wandb.log({'val_epoch_loss': val_loss, 'val_epoch_acc': val_acc})\n",
    "\n",
    "    print(f'train loss: {train_loss:.2f} | train acc: {train_acc:.2f}')\n",
    "    print(f'val   loss: {val_loss:.2f} | val   acc: {val_acc:.2f}')\n",
    "    \n",
    "    if val_loss < best_val_loss:\n",
    "        best_val_loss = val_loss\n",
    "        best_val_acc = val_acc\n",
    "        # ensure to store the model in eval mode\n",
    "        model.eval()\n",
    "        torch.save(model, 'leafsnap_model.pytorch')\n",
    "\n",
    "print(f'best val loss: {best_val_loss:.2f} | best val acc: {best_val_acc:.2f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "678e059d-f550-419d-89d4-5448fc2df5bc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test loss: 0.87 | test acc: 0.74\n"
     ]
    }
   ],
   "source": [
    "test_loss, test_acc = evaluate(model, test_dataloader, loss_func, train_mode=False)\n",
    "print(f'test loss: {test_loss:.2f} | test acc: {test_acc:.2f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "9cfd4599-7085-4d8f-a5d0-ace493ec7cca",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load model from disk an convert to onnx\n",
    "loaded_model = torch.load('leafsnap_model.pytorch')\n",
    "loaded_model.eval()\n",
    "# export needs example input\n",
    "x = next(iter(train_dataloader))[0].to(device)\n",
    "\n",
    "torch.onnx.export(loaded_model, x, 'leafsnap_model.onnx', opset_version=11,\n",
    "                  export_params=True, input_names=['input'], output_names=['output'],\n",
    "                  dynamic_axes={'input': {0: 'batch_size'},\n",
    "                                'output': {0: 'batch_size'}})"
   ]
  }
 ],
 "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": "3.9.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}