assets/hub/pytorch_vision_alexnet.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "41437a01",
   "metadata": {},
   "source": [
    "### This notebook is optionally accelerated with a GPU runtime.\n",
    "### If you would like to use this acceleration, please select the menu option \"Runtime\" -> \"Change runtime type\", select \"Hardware Accelerator\" -> \"GPU\" and click \"SAVE\"\n",
    "\n",
    "----------------------------------------------------------------------\n",
    "\n",
    "# AlexNet\n",
    "\n",
    "*Author: Pytorch Team*\n",
    "\n",
    "**The 2012 ImageNet winner achieved a top-5 error of 15.3%, more than 10.8 percentage points lower than that of the runner up.**\n",
    "\n",
    "_ | _\n",
    "- | -\n",
    "![alt](https://pytorch.org/assets/images/alexnet1.png) | ![alt](https://pytorch.org/assets/images/alexnet2.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9bb64902",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "model = torch.hub.load('pytorch/vision:v0.10.0', 'alexnet', pretrained=True)\n",
    "model.eval()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2526fd42",
   "metadata": {},
   "source": [
    "All pre-trained models expect input images normalized in the same way,\n",
    "i.e. mini-batches of 3-channel RGB images of shape `(3 x H x W)`, where `H` and `W` are expected to be at least `224`.\n",
    "The images have to be loaded in to a range of `[0, 1]` and then normalized using `mean = [0.485, 0.456, 0.406]`\n",
    "and `std = [0.229, 0.224, 0.225]`.\n",
    "\n",
    "Here's a sample execution."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "521a9830",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Download an example image from the pytorch website\n",
    "import urllib\n",
    "url, filename = (\"https://github.com/pytorch/hub/raw/master/images/dog.jpg\", \"dog.jpg\")\n",
    "try: urllib.URLopener().retrieve(url, filename)\n",
    "except: urllib.request.urlretrieve(url, filename)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4350206d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# sample execution (requires torchvision)\n",
    "from PIL import Image\n",
    "from torchvision import transforms\n",
    "input_image = Image.open(filename)\n",
    "preprocess = transforms.Compose([\n",
    "    transforms.Resize(256),\n",
    "    transforms.CenterCrop(224),\n",
    "    transforms.ToTensor(),\n",
    "    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),\n",
    "])\n",
    "input_tensor = preprocess(input_image)\n",
    "input_batch = input_tensor.unsqueeze(0) # create a mini-batch as expected by the model\n",
    "\n",
    "# move the input and model to GPU for speed if available\n",
    "if torch.cuda.is_available():\n",
    "    input_batch = input_batch.to('cuda')\n",
    "    model.to('cuda')\n",
    "\n",
    "with torch.no_grad():\n",
    "    output = model(input_batch)\n",
    "# Tensor of shape 1000, with confidence scores over ImageNet's 1000 classes\n",
    "print(output[0])\n",
    "# The output has unnormalized scores. To get probabilities, you can run a softmax on it.\n",
    "probabilities = torch.nn.functional.softmax(output[0], dim=0)\n",
    "print(probabilities)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a0d9af65",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Download ImageNet labels\n",
    "!wget https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "defe7bf0",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read the categories\n",
    "with open(\"imagenet_classes.txt\", \"r\") as f:\n",
    "    categories = [s.strip() for s in f.readlines()]\n",
    "# Show top categories per image\n",
    "top5_prob, top5_catid = torch.topk(probabilities, 5)\n",
    "for i in range(top5_prob.size(0)):\n",
    "    print(categories[top5_catid[i]], top5_prob[i].item())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1bd503be",
   "metadata": {},
   "source": [
    "### Model Description\n",
    "\n",
    "AlexNet competed in the ImageNet Large Scale Visual Recognition Challenge on September 30, 2012. The network achieved a top-5 error of 15.3%, more than 10.8 percentage points lower than that of the runner up. The original paper's primary result was that the depth of the model was essential for its high performance, which was computationally expensive, but made feasible due to the utilization of graphics processing units (GPUs) during training.\n",
    "\n",
    "The 1-crop error rates on the ImageNet dataset with the pretrained model are listed below.\n",
    "\n",
    "| Model structure | Top-1 error | Top-5 error |\n",
    "| --------------- | ----------- | ----------- |\n",
    "|  AlexNet        | 43.45       | 20.91       |\n",
    "\n",
    "### References\n",
    "\n",
    "1. [One weird trick for parallelizing convolutional neural networks](https://arxiv.org/abs/1404.5997)."
   ]
  }
 ],
 "metadata": {},
 "nbformat": 4,
 "nbformat_minor": 5
}