Buckets:
| <meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Depth estimation","local":"depth-estimation","sections":[],"depth":1}"> | |
| <link href="/docs/datasets/main/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/entry/start.4d44eea4.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/chunks/scheduler.bdbef820.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/chunks/singletons.36b689ad.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/chunks/index.8a885b74.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/chunks/paths.27092e28.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/entry/app.d83067e8.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/chunks/index.c0aea24a.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/nodes/0.bfb01985.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/chunks/each.e59479a4.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/nodes/17.0b16a5f3.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/chunks/CodeBlock.6ccca92e.js"> | |
| <link rel="modulepreload" href="/docs/datasets/main/en/_app/immutable/chunks/EditOnGithub.725ee0c1.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{"title":"Depth estimation","local":"depth-estimation","sections":[],"depth":1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="depth-estimation" 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="#depth-estimation"><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>Depth estimation</span></h1> <p data-svelte-h="svelte-10bebb7">Depth estimation datasets are used to train a model to approximate the relative distance of every pixel in an | |
| image from the camera, also known as depth. The applications enabled by these datasets primarily lie in areas like visual machine | |
| perception and perception in robotics. Example applications include mapping streets for self-driving cars. This guide will show you how to apply transformations | |
| to a depth estimation dataset.</p> <p data-svelte-h="svelte-1e3jamv">Before you start, make sure you have up-to-date versions of <code>albumentations</code> installed:</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 -->pip install -U albumentations <!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-180f5nb"><a href="https://albumentations.ai/" rel="nofollow">Albumentations</a> is a Python library for performing data augmentation | |
| for computer vision. It supports various computer vision tasks such as image classification, object | |
| detection, segmentation, and keypoint estimation.</p> <p data-svelte-h="svelte-rbl7ww">This guide uses the <a href="https://huggingface.co/datasets/sayakpaul/nyu_depth_v2" rel="nofollow">NYU Depth V2</a> dataset which is | |
| comprised of video sequences from various indoor scenes, recorded by RGB and depth cameras. The dataset consists of scenes from 3 cities and provides images along with | |
| their depth maps as labels.</p> <p data-svelte-h="svelte-pju1x0">Load the <code>train</code> split of the dataset and take a look at an example:</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-meta">>>> </span><span class="hljs-keyword">from</span> datasets <span class="hljs-keyword">import</span> load_dataset | |
| <span class="hljs-meta">>>> </span>train_dataset = load_dataset(<span class="hljs-string">"sayakpaul/nyu_depth_v2"</span>, split=<span class="hljs-string">"train"</span>) | |
| <span class="hljs-meta">>>> </span>index = <span class="hljs-number">17</span> | |
| <span class="hljs-meta">>>> </span>example = train_dataset[index] | |
| <span class="hljs-meta">>>> </span>example | |
| {<span class="hljs-string">'image'</span>: <PIL.PngImagePlugin.PngImageFile image mode=RGB size=640x480>, | |
| <span class="hljs-string">'depth_map'</span>: <PIL.TiffImagePlugin.TiffImageFile image mode=F size=640x480>}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1wdl8tg">The dataset has two fields:</p> <ul data-svelte-h="svelte-18g7ww2"><li><code>image</code>: a PIL PNG image object with <code>uint8</code> data type.</li> <li><code>depth_map</code>: a PIL Tiff image object with <code>float32</code> data type which is the depth map of the image.</li></ul> <p data-svelte-h="svelte-lgny84">It is mention-worthy that JPEG/PNG format can only store <code>uint8</code> or <code>uint16</code> data. As the depth map is <code>float32</code> data, it can’t be stored using PNG/JPEG. However, we can save the depth map using TIFF format as it supports a wider range of data types, including <code>float32</code> data.</p> <p data-svelte-h="svelte-15dthpv">Next, check out an image with:</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-meta">>>> </span>example[<span class="hljs-string">"image"</span>]<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-1f5za0g"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/depth_est_sample.png"></div> <p data-svelte-h="svelte-1rod9tf">Before we look at the depth map, we need to first convert its data type to <code>uint8</code> using <code>.convert('RGB')</code> as PIL can’t display <code>float32</code> images. Now take a look at its corresponding depth map:</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-meta">>>> </span>example[<span class="hljs-string">"depth_map"</span>].convert(<span class="hljs-string">"RGB"</span>)<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-130rqhd"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/depth_est_target.png"></div> <p data-svelte-h="svelte-16bopxj">It’s all black! You’ll need to add some color to the depth map to visualize it properly. To do that, either we can apply color automatically during display using <code>plt.imshow()</code> or create a colored depth map using <code>plt.cm</code> and then display it. In this example, we have used the latter one, as we can save/write the colored depth map later. (the utility below is taken from the <a href="https://github.com/dwofk/fast-depth/blob/master/utils.py" rel="nofollow">FastDepth repository</a>).</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-meta">>>> </span><span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np | |
| <span class="hljs-meta">>>> </span><span class="hljs-keyword">import</span> matplotlib.pyplot <span class="hljs-keyword">as</span> plt | |
| <span class="hljs-meta">>>> </span>cmap = plt.cm.viridis | |
| <span class="hljs-meta">>>> </span><span class="hljs-keyword">def</span> <span class="hljs-title function_">colored_depthmap</span>(<span class="hljs-params">depth, d_min=<span class="hljs-literal">None</span>, d_max=<span class="hljs-literal">None</span></span>): | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">if</span> d_min <span class="hljs-keyword">is</span> <span class="hljs-literal">None</span>: | |
| <span class="hljs-meta">... </span> d_min = np.<span class="hljs-built_in">min</span>(depth) | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">if</span> d_max <span class="hljs-keyword">is</span> <span class="hljs-literal">None</span>: | |
| <span class="hljs-meta">... </span> d_max = np.<span class="hljs-built_in">max</span>(depth) | |
| <span class="hljs-meta">... </span> depth_relative = (depth - d_min) / (d_max - d_min) | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">return</span> <span class="hljs-number">255</span> * cmap(depth_relative)[:,:,:<span class="hljs-number">3</span>] | |
| <span class="hljs-meta">>>> </span><span class="hljs-keyword">def</span> <span class="hljs-title function_">show_depthmap</span>(<span class="hljs-params">depth_map</span>): | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">if</span> <span class="hljs-keyword">not</span> <span class="hljs-built_in">isinstance</span>(depth_map, np.ndarray): | |
| <span class="hljs-meta">... </span> depth_map = np.array(depth_map) | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">if</span> depth_map.ndim == <span class="hljs-number">3</span>: | |
| <span class="hljs-meta">... </span> depth_map = depth_map.squeeze() | |
| <span class="hljs-meta">... </span> d_min = np.<span class="hljs-built_in">min</span>(depth_map) | |
| <span class="hljs-meta">... </span> d_max = np.<span class="hljs-built_in">max</span>(depth_map) | |
| <span class="hljs-meta">... </span> depth_map = colored_depthmap(depth_map, d_min, d_max) | |
| <span class="hljs-meta">... </span> plt.imshow(depth_map.astype(<span class="hljs-string">"uint8"</span>)) | |
| <span class="hljs-meta">... </span> plt.axis(<span class="hljs-string">"off"</span>) | |
| <span class="hljs-meta">... </span> plt.show() | |
| <span class="hljs-meta">>>> </span>show_depthmap(example[<span class="hljs-string">"depth_map"</span>])<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-16ha3wj"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/depth_est_target_viz.png"></div> <p data-svelte-h="svelte-llua6z">You can also visualize several different images and their corresponding depth maps.</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-meta">>>> </span><span class="hljs-keyword">def</span> <span class="hljs-title function_">merge_into_row</span>(<span class="hljs-params">input_image, depth_target</span>): | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">if</span> <span class="hljs-keyword">not</span> <span class="hljs-built_in">isinstance</span>(input_image, np.ndarray): | |
| <span class="hljs-meta">... </span> input_image = np.array(input_image) | |
| ... | |
| <span class="hljs-meta">... </span> d_min = np.<span class="hljs-built_in">min</span>(depth_target) | |
| <span class="hljs-meta">... </span> d_max = np.<span class="hljs-built_in">max</span>(depth_target) | |
| <span class="hljs-meta">... </span> depth_target_col = colored_depthmap(depth_target, d_min, d_max) | |
| <span class="hljs-meta">... </span> img_merge = np.hstack([input_image, depth_target_col]) | |
| ... | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">return</span> img_merge | |
| <span class="hljs-meta">>>> </span>random_indices = np.random.choice(<span class="hljs-built_in">len</span>(train_dataset), <span class="hljs-number">9</span>).tolist() | |
| <span class="hljs-meta">>>> </span>plt.figure(figsize=(<span class="hljs-number">15</span>, <span class="hljs-number">6</span>)) | |
| <span class="hljs-meta">>>> </span><span class="hljs-keyword">for</span> i, idx <span class="hljs-keyword">in</span> <span class="hljs-built_in">enumerate</span>(random_indices): | |
| <span class="hljs-meta">... </span> example = train_dataset[idx] | |
| <span class="hljs-meta">... </span> ax = plt.subplot(<span class="hljs-number">3</span>, <span class="hljs-number">3</span>, i + <span class="hljs-number">1</span>) | |
| <span class="hljs-meta">... </span> image_viz = merge_into_row( | |
| <span class="hljs-meta">... </span> example[<span class="hljs-string">"image"</span>], example[<span class="hljs-string">"depth_map"</span>] | |
| <span class="hljs-meta">... </span> ) | |
| <span class="hljs-meta">... </span> plt.imshow(image_viz.astype(<span class="hljs-string">"uint8"</span>)) | |
| <span class="hljs-meta">... </span> plt.axis(<span class="hljs-string">"off"</span>)<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-16sg8kz"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/depth_est_collage.png"></div> <p data-svelte-h="svelte-18gbq1z">Now apply some augmentations with <code>albumentations</code>. The augmentation transformations include:</p> <ul data-svelte-h="svelte-1r8fe9e"><li>Random horizontal flipping</li> <li>Random cropping</li> <li>Random brightness and contrast</li> <li>Random gamma correction</li> <li>Random hue saturation</li></ul> <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-meta">>>> </span><span class="hljs-keyword">import</span> albumentations <span class="hljs-keyword">as</span> A | |
| <span class="hljs-meta">>>> </span>crop_size = (<span class="hljs-number">448</span>, <span class="hljs-number">576</span>) | |
| <span class="hljs-meta">>>> </span>transforms = [ | |
| <span class="hljs-meta">... </span> A.HorizontalFlip(p=<span class="hljs-number">0.5</span>), | |
| <span class="hljs-meta">... </span> A.RandomCrop(crop_size[<span class="hljs-number">0</span>], crop_size[<span class="hljs-number">1</span>]), | |
| <span class="hljs-meta">... </span> A.RandomBrightnessContrast(), | |
| <span class="hljs-meta">... </span> A.RandomGamma(), | |
| <span class="hljs-meta">... </span> A.HueSaturationValue() | |
| <span class="hljs-meta">... </span>]<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1yx9z74">Additionally, define a mapping to better reflect the target key name.</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-meta">>>> </span>additional_targets = {<span class="hljs-string">"depth"</span>: <span class="hljs-string">"mask"</span>} | |
| <span class="hljs-meta">>>> </span>aug = A.Compose(transforms=transforms, additional_targets=additional_targets)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1obm0dg">With <code>additional_targets</code> defined, you can pass the target depth maps to the <code>depth</code> argument of <code>aug</code> instead of <code>mask</code>. You’ll notice this change | |
| in the <code>apply_transforms()</code> function defined below.</p> <p data-svelte-h="svelte-11hy1qw">Create a function to apply the transformation to the images as well as their depth maps:</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-meta">>>> </span><span class="hljs-keyword">def</span> <span class="hljs-title function_">apply_transforms</span>(<span class="hljs-params">examples</span>): | |
| <span class="hljs-meta">... </span> transformed_images, transformed_maps = [], [] | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">for</span> image, depth_map <span class="hljs-keyword">in</span> <span class="hljs-built_in">zip</span>(examples[<span class="hljs-string">"image"</span>], examples[<span class="hljs-string">"depth_map"</span>]): | |
| <span class="hljs-meta">... </span> image, depth_map = np.array(image), np.array(depth_map) | |
| <span class="hljs-meta">... </span> transformed = aug(image=image, depth=depth_map) | |
| <span class="hljs-meta">... </span> transformed_images.append(transformed[<span class="hljs-string">"image"</span>]) | |
| <span class="hljs-meta">... </span> transformed_maps.append(transformed[<span class="hljs-string">"depth"</span>]) | |
| ... | |
| <span class="hljs-meta">... </span> examples[<span class="hljs-string">"pixel_values"</span>] = transformed_images | |
| <span class="hljs-meta">... </span> examples[<span class="hljs-string">"labels"</span>] = transformed_maps | |
| <span class="hljs-meta">... </span> <span class="hljs-keyword">return</span> examples<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1x4vpj6">Use the <a href="/docs/datasets/main/en/package_reference/main_classes#datasets.Dataset.set_transform">set_transform()</a> function to apply the transformation on-the-fly to batches of the dataset to consume less disk space:</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-meta">>>> </span>train_dataset.set_transform(apply_transforms)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-k3k6ue">You can verify the transformation worked by indexing into the <code>pixel_values</code> and <code>labels</code> of an example image:</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-meta">>>> </span>example = train_dataset[index] | |
| <span class="hljs-meta">>>> </span>plt.imshow(example[<span class="hljs-string">"pixel_values"</span>]) | |
| <span class="hljs-meta">>>> </span>plt.axis(<span class="hljs-string">"off"</span>) | |
| <span class="hljs-meta">>>> </span>plt.show()<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-1vylqwk"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/depth_est_sample_aug.png"></div> <p data-svelte-h="svelte-1yo7m5l">Visualize the same transformation on the image’s corresponding depth map:</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-meta">>>> </span>show_depthmap(example[<span class="hljs-string">"labels"</span>])<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-1wigtnh"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/depth_est_target_aug.png"></div> <p data-svelte-h="svelte-wsimbk">You can also visualize multiple training samples reusing the previous <code>random_indices</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-meta">>>> </span>plt.figure(figsize=(<span class="hljs-number">15</span>, <span class="hljs-number">6</span>)) | |
| <span class="hljs-meta">>>> </span><span class="hljs-keyword">for</span> i, idx <span class="hljs-keyword">in</span> <span class="hljs-built_in">enumerate</span>(random_indices): | |
| <span class="hljs-meta">... </span> ax = plt.subplot(<span class="hljs-number">3</span>, <span class="hljs-number">3</span>, i + <span class="hljs-number">1</span>) | |
| <span class="hljs-meta">... </span> example = train_dataset[idx] | |
| <span class="hljs-meta">... </span> image_viz = merge_into_row( | |
| <span class="hljs-meta">... </span> example[<span class="hljs-string">"pixel_values"</span>], example[<span class="hljs-string">"labels"</span>] | |
| <span class="hljs-meta">... </span> ) | |
| <span class="hljs-meta">... </span> plt.imshow(image_viz.astype(<span class="hljs-string">"uint8"</span>)) | |
| <span class="hljs-meta">... </span> plt.axis(<span class="hljs-string">"off"</span>)<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-9icv47"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/depth_est_aug_collage.png"></div> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/datasets/blob/main/docs/source/depth_estimation.mdx" target="_blank"><span data-svelte-h="svelte-1kd6by1"><</span> <span data-svelte-h="svelte-x0xyl0">></span> <span data-svelte-h="svelte-1dajgef"><span class="underline ml-1.5">Update</span> on GitHub</span></a> <p></p> | |
| <script> | |
| { | |
| __sveltekit_w3org2 = { | |
| assets: "/docs/datasets/main/en", | |
| base: "/docs/datasets/main/en", | |
| env: {} | |
| }; | |
| const element = document.currentScript.parentElement; | |
| const data = [null,null]; | |
| Promise.all([ | |
| import("/docs/datasets/main/en/_app/immutable/entry/start.4d44eea4.js"), | |
| import("/docs/datasets/main/en/_app/immutable/entry/app.d83067e8.js") | |
| ]).then(([kit, app]) => { | |
| kit.start(app, element, { | |
| node_ids: [0, 17], | |
| data, | |
| form: null, | |
| error: null | |
| }); | |
| }); | |
| } | |
| </script> | |
Xet Storage Details
- Size:
- 36.1 kB
- Xet hash:
- 400b4dc63df9350c73a1743f9116442208999aaa0a55f9b0f6dfdd914d2e52ba
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.