transformers/main/en/tasks/image_feature_extraction.html

<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Image Feature Extraction&quot;,&quot;local&quot;:&quot;image-feature-extraction&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Image Similarity using image-feature-extraction Pipeline&quot;,&quot;local&quot;:&quot;image-similarity-using-image-feature-extraction-pipeline&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Getting Features and Similarities using AutoModel&quot;,&quot;local&quot;:&quot;getting-features-and-similarities-using-automodel&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
		<link href="/docs/transformers/main/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/entry/start.2135b7e6.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/scheduler.25b97de1.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/singletons.0f2b7d5f.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/index.e188933d.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/paths.3d04d2c6.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/entry/app.24372c84.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/index.d9030fc9.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/nodes/0.026d2fdd.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/each.e59479a4.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/nodes/405.bb6246f4.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/CodeBlock.e6cd0d95.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/DocNotebookDropdown.5ea6cb78.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/globals.7f7f1b26.js">
		<link rel="modulepreload" href="/docs/transformers/main/en/_app/immutable/chunks/EditOnGithub.91d95064.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Image Feature Extraction&quot;,&quot;local&quot;:&quot;image-feature-extraction&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Image Similarity using image-feature-extraction Pipeline&quot;,&quot;local&quot;:&quot;image-similarity-using-image-feature-extraction-pipeline&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Getting Features and Similarities using AutoModel&quot;,&quot;local&quot;:&quot;getting-features-and-similarities-using-automodel&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END -->      <p></p>   <h1 class="relative group"><a id="image-feature-extraction" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#image-feature-extraction"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Image Feature Extraction</span></h1>  <div class="flex space-x-1 absolute z-10 right-0 top-0"> <div class="relative colab-dropdown "> <button class="  " type="button">  <img alt="Open In Colab" class="!m-0" src="https://colab.research.google.com/assets/colab-badge.svg"> </button>    </div> <div class="relative colab-dropdown "> <button class="  " type="button">  <img alt="Open In Studio Lab" class="!m-0" src="https://studiolab.sagemaker.aws/studiolab.svg"> </button>    </div></div> <p data-svelte-h="svelte-htpxsg">Image feature extraction is the task of extracting semantically meaningful features given an image. This has many use cases, including image similarity and image retrieval. Moreover, most computer vision models can be used for image feature extraction, where one can remove the task-specific head (image classification, object detection etc) and get the features. These features are very useful on a higher level: edge detection, corner detection and so on. They may also contain information about the real world (e.g. what a cat looks like) depending on how deep the model is. Therefore, these outputs can be used to train new classifiers on a specific dataset.</p> <p data-svelte-h="svelte-dkb44i">In this guide, you will:</p> <ul data-svelte-h="svelte-o9mygf"><li>Learn to build a simple image similarity system on top of the <code>image-feature-extraction</code> pipeline.</li> <li>Accomplish the same task with bare model inference.</li></ul>  <h2 class="relative group"><a id="image-similarity-using-image-feature-extraction-pipeline" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#image-similarity-using-image-feature-extraction-pipeline"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Image Similarity using image-feature-extraction Pipeline</span></h2> <p data-svelte-h="svelte-1y1lzmz">We have two images of cats sitting on top of fish nets, one of them is generated.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> PIL <span class="hljs-keyword">import</span> Image
<span class="hljs-keyword">import</span> requests

img_urls = [<span class="hljs-string">&quot;https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/cats.png&quot;</span>, <span class="hljs-string">&quot;https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/cats.jpeg&quot;</span>]
image_real = Image.<span class="hljs-built_in">open</span>(requests.get(img_urls[<span class="hljs-number">0</span>], stream=<span class="hljs-literal">True</span>).raw).convert(<span class="hljs-string">&quot;RGB&quot;</span>)
image_gen = Image.<span class="hljs-built_in">open</span>(requests.get(img_urls[<span class="hljs-number">1</span>], stream=<span class="hljs-literal">True</span>).raw).convert(<span class="hljs-string">&quot;RGB&quot;</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ufkt7t">Let’s see the pipeline in action. First, initialize the pipeline. If you don’t pass any model to it, the pipeline will be automatically initialized with <a href="google/vit-base-patch16-224">google/vit-base-patch16-224</a>. If you’d like to calculate similarity, set <code>pool</code> to True.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">import</span> torch
<span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> pipeline

DEVICE = torch.device(<span class="hljs-string">&#x27;cuda&#x27;</span> <span class="hljs-keyword">if</span> torch.cuda.is_available() <span class="hljs-keyword">else</span> <span class="hljs-string">&#x27;cpu&#x27;</span>)
pipe = pipeline(task=<span class="hljs-string">&quot;image-feature-extraction&quot;</span>, model_name=<span class="hljs-string">&quot;google/vit-base-patch16-384&quot;</span>, device=DEVICE, pool=<span class="hljs-literal">True</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1wqrlcg">To infer with <code>pipe</code> pass both images to it.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->outputs = pipe([image_real, image_gen])<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ejn9fo">The output contains pooled embeddings of those two images.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-comment"># get the length of a single output</span>
<span class="hljs-built_in">print</span>(<span class="hljs-built_in">len</span>(outputs[<span class="hljs-number">0</span>][<span class="hljs-number">0</span>]))
<span class="hljs-comment"># show outputs</span>
<span class="hljs-built_in">print</span>(outputs)

<span class="hljs-comment"># 768</span>
<span class="hljs-comment"># [[[-0.03909236937761307, 0.43381670117378235, -0.06913255900144577,</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-7ov6zk">To get the similarity score, we need to pass them to a similarity function.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> torch.nn.functional <span class="hljs-keyword">import</span> cosine_similarity

similarity_score = cosine_similarity(torch.Tensor(outputs[<span class="hljs-number">0</span>]),
                                     torch.Tensor(outputs[<span class="hljs-number">1</span>]), dim=<span class="hljs-number">1</span>)

<span class="hljs-built_in">print</span>(similarity_score)

<span class="hljs-comment"># tensor([0.6043])</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1t8cfg9">If you want to get the last hidden states before pooling, avoid passing any value for the <code>pool</code> parameter, as it is set to <code>False</code> by default. These hidden states are useful for training new classifiers or models based on the features from the model.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->pipe = pipeline(task=<span class="hljs-string">&quot;image-feature-extraction&quot;</span>, model_name=<span class="hljs-string">&quot;google/vit-base-patch16-224&quot;</span>, device=DEVICE)
output = pipe(image_real)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-u3d9d7">Since the outputs are unpooled, we get the last hidden states where the first dimension is the batch size, and the last two are the embedding shape.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np
<span class="hljs-built_in">print</span>(np.array(outputs).shape)
<span class="hljs-comment"># (1, 197, 768)</span><!-- HTML_TAG_END --></pre></div>  <h2 class="relative group"><a id="getting-features-and-similarities-using-automodel" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#getting-features-and-similarities-using-automodel"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Getting Features and Similarities using AutoModel</span></h2> <p data-svelte-h="svelte-18d7a3m">We can also use <code>AutoModel</code> class of transformers to get the features. <code>AutoModel</code> loads any transformers model with no task-specific head, and we can use this to get the features.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> AutoImageProcessor, AutoModel

processor = AutoImageProcessor.from_pretrained(<span class="hljs-string">&quot;google/vit-base-patch16-224&quot;</span>)
model = AutoModel.from_pretrained(<span class="hljs-string">&quot;google/vit-base-patch16-224&quot;</span>).to(DEVICE)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1mcm377">Let’s write a simple function for inference. We will pass the inputs to the <code>processor</code> first and pass its outputs to the <code>model</code>.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">infer</span>(<span class="hljs-params">image</span>):
  inputs = processor(image, return_tensors=<span class="hljs-string">&quot;pt&quot;</span>).to(DEVICE)
  outputs = model(**inputs)
  <span class="hljs-keyword">return</span> outputs.pooler_output<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1ej6c6d">We can pass the images directly to this function and get the embeddings.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START -->embed_real = infer(image_real)
embed_gen = infer(image_gen)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1t2j36n">We can get the similarity again over the embeddings.</p> <div class="code-block relative"><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5   text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">from</span> torch.nn.functional <span class="hljs-keyword">import</span> cosine_similarity

similarity_score = cosine_similarity(embed_real, embed_gen, dim=<span class="hljs-number">1</span>)
<span class="hljs-built_in">print</span>(similarity_score)

<span class="hljs-comment"># tensor([0.6061], device=&#x27;cuda:0&#x27;, grad_fn=&lt;SumBackward1&gt;)</span><!-- HTML_TAG_END --></pre></div> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/transformers/blob/main/docs/source/en/tasks/image_feature_extraction.md" target="_blank"><span data-svelte-h="svelte-1kd6by1">&lt;</span> <span data-svelte-h="svelte-x0xyl0">&gt;</span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a>  <p></p> 
			
			<script>
				{
					__sveltekit_1xexzbk = {
						assets: "/docs/transformers/main/en",
						base: "/docs/transformers/main/en",
						env: {}
					};

					const element = document.currentScript.parentElement;

					const data = [null,null];

					Promise.all([
						import("/docs/transformers/main/en/_app/immutable/entry/start.2135b7e6.js"),
						import("/docs/transformers/main/en/_app/immutable/entry/app.24372c84.js")
					]).then(([kit, app]) => {
						kit.start(app, element, {
							node_ids: [0, 405],
							data,
							form: null,
							error: null
						});
					});
				}
			</script>