{"guide": {"name": "image-classification-in-pytorch", "category": "other-tutorials", "pretty_category": "Other Tutorials", "guide_index": null, "absolute_index": 66, "pretty_name": "Image Classification In Pytorch", "content": "# Image Classification in PyTorch\n\n\n\n\n## Introduction\n\nImage classification is a central task in computer vision. Building better classifiers to classify what object is present in a picture is an active area of research, as it has applications stretching from autonomous vehicles to medical imaging.\n\nSuch models are perfect to use with Gradio's _image_ input component, so in this tutorial we will build a web demo to classify images using Gradio. We will be able to build the whole web application in Python, and it will look like the demo on the bottom of the page.\n\nLet's get started!\n\n### Prerequisites\n\nMake sure you have the `gradio` Python package already [installed](/getting_started). We will be using a pretrained image classification model, so you should also have `torch` installed.\n\n## Step 1 \u2014 Setting up the Image Classification Model\n\nFirst, we will need an image classification model. For this tutorial, we will use a pretrained Resnet-18 model, as it is easily downloadable from [PyTorch Hub](https://pytorch.org/hub/pytorch_vision_resnet/). You can use a different pretrained model or train your own.\n\n```python\nimport torch\n\nmodel = torch.hub.load('pytorch/vision:v0.6.0', 'resnet18', pretrained=True).eval()\n```\n\nBecause we will be using the model for inference, we have called the `.eval()` method.\n\n## Step 2 \u2014 Defining a `predict` function\n\nNext, we will need to define a function that takes in the _user input_, which in this case is an image, and returns the prediction. The prediction should be returned as a dictionary whose keys are class name and values are confidence probabilities. We will load the class names from this [text file](https://git.io/JJkYN).\n\nIn the case of our pretrained model, it will look like this:\n\n```python\nimport requests\nfrom PIL import Image\nfrom torchvision import transforms\n\n# Download human-readable labels for ImageNet.\nresponse = requests.get(\"https://git.io/JJkYN\")\nlabels = response.text.split(\"\\n\")\n\ndef predict(inp):\n  inp = transforms.ToTensor()(inp).unsqueeze(0)\n  with torch.no_grad():\n    prediction = torch.nn.functional.softmax(model(inp)[0], dim=0)\n    confidences = {labels[i]: float(prediction[i]) for i in range(1000)}\n  return confidences\n```\n\nLet's break this down. The function takes one parameter:\n\n- `inp`: the input image as a `PIL` image\n\nThen, the function converts the image to a PIL Image and then eventually a PyTorch `tensor`, passes it through the model, and returns:\n\n- `confidences`: the predictions, as a dictionary whose keys are class labels and whose values are confidence probabilities\n\n## Step 3 \u2014 Creating a Gradio Interface\n\nNow that we have our predictive function set up, we can create a Gradio Interface around it.\n\nIn this case, the input component is a drag-and-drop image component. To create this input, we use `Image(type=\"pil\")` which creates the component and handles the preprocessing to convert that to a `PIL` image.\n\nThe output component will be a `Label`, which displays the top labels in a nice form. Since we don't want to show all 1,000 class labels, we will customize it to show only the top 3 images by constructing it as `Label(num_top_classes=3)`.\n\nFinally, we'll add one more parameter, the `examples`, which allows us to prepopulate our interfaces with a few predefined examples. The code for Gradio looks like this:\n\n```python\nimport gradio as gr\n\ngr.Interface(fn=predict,\n             inputs=gr.Image(type=\"pil\"),\n             outputs=gr.Label(num_top_classes=3),\n             examples=[\"lion.jpg\", \"cheetah.jpg\"]).launch()\n```\n\nThis produces the following interface, which you can try right here in your browser (try uploading your own examples!):\n\n<gradio-app space=\"gradio/pytorch-image-classifier\">\n\n\n---\n\nAnd you're done! That's all the code you need to build a web demo for an image classifier. If you'd like to share with others, try setting `share=True` when you `launch()` the Interface!\n", "tags": ["VISION", "RESNET", "PYTORCH"], "spaces": ["https://huggingface.co/spaces/abidlabs/pytorch-image-classifier", "https://huggingface.co/spaces/pytorch/ResNet", "https://huggingface.co/spaces/pytorch/ResNext", "https://huggingface.co/spaces/pytorch/SqueezeNet"], "url": "/guides/image-classification-in-pytorch/", "contributor": null}}