inference.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_312819/875621319.py:45: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.\n",
      "  checkpoint = torch.load(checkpoint_path)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predictions:\n",
      "n02105412: 26.26%\n",
      "n02106662: 14.42%\n",
      "n02099601: 8.62%\n",
      "n02105162: 7.17%\n",
      "n02105251: 3.21%\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(tensor([0.2626, 0.1442, 0.0862, 0.0717, 0.0321], device='cuda:0'),\n",
       " tensor([227, 235, 207, 225, 226], device='cuda:0'))"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import torch\n",
    "import torchvision.transforms as transforms\n",
    "from torchvision.datasets import ImageFolder\n",
    "from PIL import Image\n",
    "import os\n",
    "\n",
    "# Define the device\n",
    "device = (\n",
    "    \"cuda\"\n",
    "    if torch.cuda.is_available()\n",
    "    else \"mps\"\n",
    "    if torch.backends.mps.is_available()\n",
    "    else \"cpu\"\n",
    ")\n",
    "\n",
    "class Params:\n",
    "    def __init__(self):\n",
    "        self.batch_size = 512\n",
    "        self.name = \"resnet_50\"\n",
    "        self.workers = 16\n",
    "        self.lr = 0.1\n",
    "        self.momentum = 0.9\n",
    "        self.weight_decay = 1e-4\n",
    "        self.lr_step_size = 30\n",
    "        self.lr_gamma = 0.1\n",
    "\n",
    "    def __repr__(self):\n",
    "        return str(self.__dict__)\n",
    "    \n",
    "    def __eq__(self, other):\n",
    "        return self.__dict__ == other.__dict__\n",
    "    \n",
    "params = Params()\n",
    "\n",
    "# Path to the saved model checkpoint\n",
    "checkpoint_path = \"checkpoints/resnet_50/checkpoint.pth\"\n",
    "\n",
    "# Load the model architecture\n",
    "from model import ResNet50  # Assuming resnet.py contains your model definition\n",
    "\n",
    "num_classes = 1000  # Adjust this to match your dataset\n",
    "model = ResNet50(num_classes=num_classes).to(device)\n",
    "\n",
    "# Load the trained model weights\n",
    "checkpoint = torch.load(checkpoint_path)\n",
    "model.load_state_dict(checkpoint[\"model\"])\n",
    "\n",
    "model.eval()\n",
    "\n",
    "# Define transformations for inference\n",
    "inference_transforms = transforms.Compose([\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Resize(size=256),\n",
    "    transforms.CenterCrop(224),\n",
    "    transforms.Normalize(mean=[0.485, 0.485, 0.406], std=[0.229, 0.224, 0.225]),\n",
    "])\n",
    "\n",
    "# Function to make predictions on a single image\n",
    "def predict(image_path, model, transforms, class_names=None):\n",
    "    # Load and transform the image\n",
    "    image = Image.open(image_path).convert(\"RGB\")\n",
    "    image_tensor = transforms(image).unsqueeze(0).to(device)\n",
    "\n",
    "    # Forward pass\n",
    "    with torch.no_grad():\n",
    "        output = model(image_tensor)\n",
    "        probabilities = torch.nn.functional.softmax(output[0], dim=0)\n",
    "        top_prob, top_class = probabilities.topk(5, largest=True, sorted=True)\n",
    "\n",
    "    # Display the top predictions\n",
    "    print(\"Predictions:\")\n",
    "    for i in range(top_prob.size(0)):\n",
    "        class_name = class_names[top_class[i]] if class_names else f\"Class {top_class[i].item()}\"\n",
    "        print(f\"{class_name}: {top_prob[i].item() * 100:.2f}%\")\n",
    "\n",
    "    return top_prob, top_class\n",
    "\n",
    "# Define the dataset class names if available (optional)\n",
    "# Example: Load class names from the training dataset folder\n",
    "dataset_path = 'Imagenet/ILSVRC/Data/CLS-LOC/train'\n",
    "class_names = sorted(os.listdir(dataset_path))\n",
    "\n",
    "# Path to the image for inference\n",
    "image_path = \"dog.png\"  # Replace with the actual path to your test image\n",
    "\n",
    "# Make a prediction\n",
    "predict(image_path, model, inference_transforms, class_names=class_names)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predictions:\n",
      "kelpie: 26.26%\n",
      "German shepherd, German shepherd dog, German police dog, alsatian: 14.42%\n",
      "golden retriever: 8.62%\n",
      "malinois: 7.17%\n",
      "briard: 3.21%\n"
     ]
    }
   ],
   "source": [
    "# Path to the class mapping text file\n",
    "class_mapping_file = \"Imagenet/LOC_synset_mapping.txt\"  # Replace with the actual path\n",
    "\n",
    "# Read the file and create a mapping\n",
    "id_to_name = {}\n",
    "with open(class_mapping_file, 'r') as f:\n",
    "    for line in f:\n",
    "        class_id, class_name = line.strip().split(\" \", 1)\n",
    "        id_to_name[class_id] = class_name\n",
    "\n",
    "# Model output (example)\n",
    "predicted_ids = [\"n02105412\", \"n02106662\", \"n02099601\", \"n02105162\", \"n02105251\"]\n",
    "probabilities = [26.26, 14.42, 8.62, 7.17, 3.21]\n",
    "\n",
    "# Map the IDs to their names\n",
    "predicted_names = [id_to_name[class_id] for class_id in predicted_ids]\n",
    "\n",
    "# Display the predictions\n",
    "print(\"Predictions:\")\n",
    "for prob, name in zip(probabilities, predicted_names):\n",
    "    print(f\"{name}: {prob:.2f}%\")\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "env",
   "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": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}