Linear_Probe.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/hmin27/2023_DL_Clip/blob/main/Linear_Probe.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rx2hKkBgzUkz"
      },
      "source": [
        "# CLIP Fewshot Linear Probe\n",
        "- 1, 2, 4, 8, 16 shots\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "UU5QbpXrDWUl",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3f6be4e9-c096-4bfc-f08a-b60eed19f651"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Installing collected packages: clip\n",
            "Successfully installed clip-1.0\n"
          ]
        }
      ],
      "source": [
        "!pip install ftfy regex tqdm\n",
        "!pip install git+https://github.com/openai/CLIP.git"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 29,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wc7W2ooyDcLZ",
        "outputId": "65d31b08-52b9-4f0a-e591-65694a67ad23"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "cuda:0\n"
          ]
        }
      ],
      "source": [
        "import os\n",
        "import clip\n",
        "import torch\n",
        "from torch.utils.data import DataLoader\n",
        "from torchvision.datasets import ImageFolder\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from sklearn.metrics import accuracy_score\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "BATCH_SIZE = 8\n",
        "\n",
        "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
        "print(device)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p52DoKG6DmVA"
      },
      "source": [
        "# Prepare the Model and Data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "RnmtrDVrE9xn",
        "outputId": "356516a0-9932-4f42-bbc2-c2a06d89e2a4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ],
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "id": "X4faeiH7QkCV",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "04727d57-2ede-4492-f770-6026faafcfd1"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "100%|████████████████████████████████████████| 338M/338M [00:02<00:00, 147MiB/s]\n"
          ]
        }
      ],
      "source": [
        "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
        "model, preprocess = clip.load(\"ViT-B/32\", device=device, jit=False)\n",
        "model = model.to(torch.float32)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fL2zwBDdNvZw"
      },
      "source": [
        "# Training"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model = model.eval()\n",
        "\n",
        "# Load your few-shot data\n",
        "data_folder_base = '/content/drive/MyDrive/Study/DL_CLIP/Food_fewshot/{}_shot/{}'\n",
        "shot_list = [1, 2, 4, 8, 16]\n",
        "accuracies = []\n",
        "\n",
        "for num_shots in shot_list:\n",
        "    data_folder_train = data_folder_base.format(num_shots, 'train')\n",
        "    data_folder_test = data_folder_base.format(num_shots, 'test')\n",
        "\n",
        "    train_dataset = ImageFolder(root=data_folder_train, transform=preprocess)\n",
        "    test_dataset = ImageFolder(root=data_folder_test, transform=preprocess)\n",
        "\n",
        "    train_loader = DataLoader(train_dataset, batch_size=BATCH_SIZE, shuffle=True)\n",
        "    test_loader = DataLoader(test_dataset, batch_size=BATCH_SIZE, shuffle=False)\n",
        "\n",
        "    # Extract features from the CLIP model\n",
        "    def get_features(dataset):\n",
        "        all_features = []\n",
        "        all_labels = []\n",
        "\n",
        "        with torch.no_grad():\n",
        "            for images, labels in DataLoader(dataset, batch_size=BATCH_SIZE):\n",
        "                images = images.to(device)\n",
        "                features = model.encode_image(images)\n",
        "                all_features.append(features)\n",
        "                all_labels.append(labels)\n",
        "\n",
        "        return torch.cat(all_features).cpu().numpy(), torch.cat(all_labels).cpu().numpy()\n",
        "\n",
        "    train_features, train_labels = get_features(train_dataset)\n",
        "    test_features, test_labels = get_features(test_dataset)\n",
        "\n",
        "    # Perform logistic regression with the extracted features\n",
        "    classifier = LogisticRegression(random_state=0, C=0.316, max_iter=1000, verbose=1)\n",
        "    classifier.fit(train_features, train_labels)\n",
        "\n",
        "    # Evaluate using the logistic regression classifier\n",
        "    predictions = classifier.predict(test_features)\n",
        "    accuracy = accuracy_score(test_labels, predictions)\n",
        "    accuracies.append(accuracy)\n",
        "    print(f\"Accuracy on {num_shots}-shot data: {accuracy * 100:.2f}%\")\n",
        "\n",
        "# Plot the accuracies\n",
        "plt.plot(shot_list, accuracies, marker='o')\n",
        "for i, txt in enumerate(accuracies):\n",
        "    plt.annotate(f'{txt * 100:.2f}%', (shot_list[i], accuracies[i]), textcoords=\"offset points\", xytext=(0,10), ha='center')\n",
        "plt.title('Accuracy vs Number of Shots')\n",
        "plt.xlabel('Number of Shots')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 559
        },
        "id": "ELOeMGRK3eFd",
        "outputId": "0dab1529-5c5b-4f6a-d7c8-c736dfcef2d8"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy on 1-shot data: 54.55%\n",
            "Accuracy on 2-shot data: 75.76%\n",
            "Accuracy on 4-shot data: 84.85%\n",
            "Accuracy on 8-shot data: 93.18%\n",
            "Accuracy on 16-shot data: 93.18%\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": [],
      "gpuType": "T4",
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "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": "3.9.18"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}