Hugging Face's logo

Buckets:
hf-doc-build
/
doc-dev

Files
xet
hf-doc-build/doc-dev / transformers /main /ko /tasks /mask_generation.html
rtrm's picture
rtrm
download
raw
37.9 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;마스크 생성&quot;,&quot;local&quot;:&quot;mask-generation&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;마스크 생성 파이프라인&quot;,&quot;local&quot;:&quot;mask-generation-pipeline&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;모델 추론&quot;,&quot;local&quot;:&quot;model-inference&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;포인트 프롬프팅&quot;,&quot;local&quot;:&quot;point-prompting&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;박스 프롬프팅&quot;,&quot;local&quot;:&quot;box-prompting&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
 <link href="/docs/transformers/main/ko/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/entry/start.9aa88961.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/chunks/scheduler.9bc65507.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/chunks/singletons.9eec45c3.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/chunks/index.3b203c72.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/chunks/paths.566078f7.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/entry/app.84fb67c3.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/chunks/index.707bf1b6.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/nodes/0.1c99376b.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/chunks/each.e59479a4.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/nodes/71.e649083b.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/chunks/CodeBlock.54a9f38d.js">
 <link rel="modulepreload" href="/docs/transformers/main/ko/_app/immutable/chunks/EditOnGithub.922df6ba.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;마스크 생성&quot;,&quot;local&quot;:&quot;mask-generation&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;마스크 생성 파이프라인&quot;,&quot;local&quot;:&quot;mask-generation-pipeline&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;모델 추론&quot;,&quot;local&quot;:&quot;model-inference&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;포인트 프롬프팅&quot;,&quot;local&quot;:&quot;point-prompting&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3},{&quot;title&quot;:&quot;박스 프롬프팅&quot;,&quot;local&quot;:&quot;box-prompting&quot;,&quot;sections&quot;:[],&quot;depth&quot;:3}],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="mask-generation" 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="#mask-generation"><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>마스크 생성</span></h1> <p data-svelte-h="svelte-mmew1k">마스크 생성(Mask generation)은 이미지에 대한 의미 있는 마스크를 생성하는 작업입니다.
이 작업은 <a href="semantic_segmentation">이미지 분할</a>과 매우 유사하지만, 많은 차이점이 있습니다. 이미지 분할 모델은 라벨이 달린 데이터셋으로 학습되며, 학습 중에 본 클래스들로만 제한됩니다. 이미지가 주어지면, 이미지 분할 모델은 여러 마스크와 그에 해당하는 클래스를 반환합니다.</p> <p data-svelte-h="svelte-9b7hts">반면, 마스크 생성 모델은 대량의 데이터로 학습되며 두 가지 모드로 작동합니다.</p> <ul data-svelte-h="svelte-ryg2ha"><li>프롬프트 모드(Prompting mode): 이 모드에서는 모델이 이미지와 프롬프트를 입력받습니다. 프롬프트는 이미지 내 객체의 2D 좌표(XY 좌표)나 객체를 둘러싼 바운딩 박스가 될 수 있습니다. 프롬프트 모드에서는 모델이 프롬프트가 가리키는 객체의 마스크만 반환합니다.</li> <li>전체 분할 모드(Segment Everything mode): 이 모드에서는 주어진 이미지 내에서 모든 마스크를 생성합니다. 이를 위해 그리드 형태의 점들을 생성하고 이를 이미지에 오버레이하여 추론합니다.</li></ul> <p data-svelte-h="svelte-10uzacz">마스크 생성 작업은 <a href="model_doc/sam">전체 분할 모드(Segment Anything Model, SAM)</a>에 의해 지원됩니다. SAM은 Vision Transformer 기반 이미지 인코더, 프롬프트 인코더, 그리고 양방향 트랜스포머 마스크 디코더로 구성된 강력한 모델입니다. 이미지와 프롬프트는 인코딩되고, 디코더는 이러한 임베딩을 받아 유효한 마스크를 생성합니다.</p> <div class="flex justify-center" data-svelte-h="svelte-18bgtyc"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/sam.png" alt="SAM Architecture"></div> <p data-svelte-h="svelte-i13rxt">SAM은 대규모 데이터를 다룰 수 있는 강력한 분할 기반 모델입니다. 이 모델은 100만 개의 이미지와 11억 개의 마스크를 포함하는 <a href="https://ai.meta.com/datasets/segment-anything/" rel="nofollow">SA-1B</a> 데이터 세트로 학습되었습니다.</p> <p data-svelte-h="svelte-5anmm9">이 가이드에서는 다음과 같은 내용을 배우게 됩니다:</p> <ul data-svelte-h="svelte-34u4u8"><li>배치 처리와 함께 전체 분할 모드에서 추론하는 방법</li> <li>포인트 프롬프팅 모드에서 추론하는 방법</li> <li>박스 프롬프팅 모드에서 추론하는 방법</li></ul> <p data-svelte-h="svelte-1dx4ib6">먼저, <code>transformers</code>를 설치해 봅시다:</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 -q transformers<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="mask-generation-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="#mask-generation-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>마스크 생성 파이프라인</span></h2> <p data-svelte-h="svelte-13kgcmm">마스크 생성 모델로 추론하는 가장 쉬운 방법은 <code>mask-generation</code> 파이프라인을 사용하는 것입니다.</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">&gt;&gt;&gt; </span><span class="hljs-keyword">from</span> transformers <span class="hljs-keyword">import</span> pipeline
<span class="hljs-meta">&gt;&gt;&gt; </span>checkpoint = <span class="hljs-string">&quot;facebook/sam-vit-base&quot;</span>
<span class="hljs-meta">&gt;&gt;&gt; </span>mask_generator = pipeline(model=checkpoint, task=<span class="hljs-string">&quot;mask-generation&quot;</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-ltc9d3">이미지를 예시로 봅시다.</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_url = <span class="hljs-string">&quot;https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg&quot;</span>
image = Image.<span class="hljs-built_in">open</span>(requests.get(img_url, stream=<span class="hljs-literal">True</span>).raw).convert(<span class="hljs-string">&quot;RGB&quot;</span>)<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-5ac7qo"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg" alt="Example Image"></div> <p data-svelte-h="svelte-sso2e2">전체적으로 분할해봅시다. <code>points-per-batch</code>는 전체 분할 모드에서 점들의 병렬 추론을 가능하게 합니다. 이를 통해 추론 속도가 빨라지지만, 더 많은 메모리를 소모하게 됩니다. 또한, SAM은 이미지가 아닌 점들에 대해서만 배치 처리를 지원합니다. <code>pred_iou_thresh</code>는 IoU 신뢰 임계값으로, 이 임계값을 초과하는 마스크만 반환됩니다.</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 -->masks = mask_generator(image, points_per_batch=<span class="hljs-number">128</span>, pred_iou_thresh=<span class="hljs-number">0.88</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1iijjkz"><code>masks</code> 는 다음과 같이 생겼습니다:</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-string">&#x27;masks&#x27;</span>: [array([[False, False, False, ..., True, True, True],
 [False, False, False, ..., True, True, True],
 [False, False, False, ..., True, True, True],
 ...,
 [False, False, False, ..., False, False, False],
 [False, False, False, ..., False, False, False],
 [False, False, False, ..., False, False, False]]),
 array([[False, False, False, ..., False, False, False],
 [False, False, False, ..., False, False, False],
 [False, False, False, ..., False, False, False],
 ...,
<span class="hljs-string">&#x27;scores&#x27;</span>: tensor([0.9972, 0.9917,
 ...,
}<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-894oei">위 내용을 아래와 같이 시각화할 수 있습니다:</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> matplotlib.pyplot <span class="hljs-keyword">as</span> plt
plt.imshow(image, cmap=<span class="hljs-string">&#x27;gray&#x27;</span>)
<span class="hljs-keyword">for</span> i, mask <span class="hljs-keyword">in</span> <span class="hljs-built_in">enumerate</span>(masks[<span class="hljs-string">&quot;masks&quot;</span>]):
 plt.imshow(mask, cmap=<span class="hljs-string">&#x27;viridis&#x27;</span>, alpha=<span class="hljs-number">0.1</span>, vmin=<span class="hljs-number">0</span>, vmax=<span class="hljs-number">1</span>)
plt.axis(<span class="hljs-string">&#x27;off&#x27;</span>)
plt.show()<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1sn01nr">아래는 회색조 원본 이미지에 다채로운 색상의 맵을 겹쳐놓은 모습입니다. 매우 인상적인 결과입니다.</p> <div class="flex justify-center" data-svelte-h="svelte-1xv3qg"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee_segmented.png" alt="Visualized"></div> <h2 class="relative group"><a id="model-inference" 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="#model-inference"><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>모델 추론</span></h2> <h3 class="relative group"><a id="point-prompting" 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="#point-prompting"><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>포인트 프롬프팅</span></h3> <p data-svelte-h="svelte-fmtk55">파이프라인 없이도 모델을 사용할 수 있습니다. 이를 위해 모델과 프로세서를 초기화해야 합니다.</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> SamModel, SamProcessor
<span class="hljs-keyword">import</span> torch
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>)
model = SamModel.from_pretrained(<span class="hljs-string">&quot;facebook/sam-vit-base&quot;</span>).to(device)
processor = SamProcessor.from_pretrained(<span class="hljs-string">&quot;facebook/sam-vit-base&quot;</span>)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-w6sf9k">포인트 프롬프팅을 하기 위해, 입력 포인트를 프로세서에 전달한 다음, 프로세서 출력을 받아 모델에 전달하여 추론합니다. 모델 출력을 후처리하려면, 출력과 함께 프로세서의 초기 출력에서 가져온 <code>original_sizes</code>와 <code>reshaped_input_sizes</code>를 전달해야 합니다. 왜냐하면, 프로세서가 이미지 크기를 조정하고 출력을 추정해야 하기 때문입니다.</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 -->input_points = [[[<span class="hljs-number">2592</span>, <span class="hljs-number">1728</span>]]] <span class="hljs-comment"># 벌의 포인트 위치</span>
inputs = processor(image, input_points=input_points, return_tensors=<span class="hljs-string">&quot;pt&quot;</span>).to(device)
<span class="hljs-keyword">with</span> torch.no_grad():
 outputs = model(**inputs)
masks = processor.image_processor.post_process_masks(outputs.pred_masks.cpu(), inputs[<span class="hljs-string">&quot;original_sizes&quot;</span>].cpu(), inputs[<span class="hljs-string">&quot;reshaped_input_sizes&quot;</span>].cpu())<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-8jfq4b"><code>masks</code> 출력으로 세 가지 마스크를 시각화할 수 있습니다.</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> matplotlib.pyplot <span class="hljs-keyword">as</span> plt
<span class="hljs-keyword">import</span> numpy <span class="hljs-keyword">as</span> np
fig, axes = plt.subplots(<span class="hljs-number">1</span>, <span class="hljs-number">4</span>, figsize=(<span class="hljs-number">15</span>, <span class="hljs-number">5</span>))
axes[<span class="hljs-number">0</span>].imshow(image)
axes[<span class="hljs-number">0</span>].set_title(<span class="hljs-string">&#x27;Original Image&#x27;</span>)
mask_list = [masks[<span class="hljs-number">0</span>][<span class="hljs-number">0</span>][<span class="hljs-number">0</span>].numpy(), masks[<span class="hljs-number">0</span>][<span class="hljs-number">0</span>][<span class="hljs-number">1</span>].numpy(), masks[<span class="hljs-number">0</span>][<span class="hljs-number">0</span>][<span class="hljs-number">2</span>].numpy()]
<span class="hljs-keyword">for</span> i, mask <span class="hljs-keyword">in</span> <span class="hljs-built_in">enumerate</span>(mask_list, start=<span class="hljs-number">1</span>):
 overlayed_image = np.array(image).copy()
overlayed_image[:,:,<span class="hljs-number">0</span>] = np.where(mask == <span class="hljs-number">1</span>, <span class="hljs-number">255</span>, overlayed_image[:,:,<span class="hljs-number">0</span>])
 overlayed_image[:,:,<span class="hljs-number">1</span>] = np.where(mask == <span class="hljs-number">1</span>, <span class="hljs-number">0</span>, overlayed_image[:,:,<span class="hljs-number">1</span>])
 overlayed_image[:,:,<span class="hljs-number">2</span>] = np.where(mask == <span class="hljs-number">1</span>, <span class="hljs-number">0</span>, overlayed_image[:,:,<span class="hljs-number">2</span>])
axes[i].imshow(overlayed_image)
 axes[i].set_title(<span class="hljs-string">f&#x27;Mask <span class="hljs-subst">{i}</span>&#x27;</span>)
<span class="hljs-keyword">for</span> ax <span class="hljs-keyword">in</span> axes:
 ax.axis(<span class="hljs-string">&#x27;off&#x27;</span>)
plt.show()<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-od5m1y"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/masks.png" alt="Visualized"></div> <h3 class="relative group"><a id="box-prompting" 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="#box-prompting"><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>박스 프롬프팅</span></h3> <p data-svelte-h="svelte-1yuitoy">박스 프롬프팅도 포인트 프롬프팅과 유사한 방식으로 할 수 있습니다. 입력 박스를 <code>[x_min, y_min, x_max, y_max]</code> 형식의 리스트로 작성하여 이미지와 함께 <code>processor</code>에 전달할 수 있습니다. 프로세서 출력을 받아 모델에 직접 전달한 후, 다시 출력을 후처리해야 합니다.</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"># 벌 주위의 바운딩 박스</span>
box = [<span class="hljs-number">2350</span>, <span class="hljs-number">1600</span>, <span class="hljs-number">2850</span>, <span class="hljs-number">2100</span>]
inputs = processor(
 image,
 input_boxes=[[[box]]],
 return_tensors=<span class="hljs-string">&quot;pt&quot;</span>
 ).to(<span class="hljs-string">&quot;cuda&quot;</span>)
<span class="hljs-keyword">with</span> torch.no_grad():
 outputs = model(**inputs)
mask = processor.image_processor.post_process_masks(
 outputs.pred_masks.cpu(),
 inputs[<span class="hljs-string">&quot;original_sizes&quot;</span>].cpu(),
 inputs[<span class="hljs-string">&quot;reshaped_input_sizes&quot;</span>].cpu()
)[<span class="hljs-number">0</span>][<span class="hljs-number">0</span>][<span class="hljs-number">0</span>].numpy()<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-wkxo0l">이제 아래와 같이, 벌 주위의 바운딩 박스를 시각화할 수 있습니다.</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> matplotlib.patches <span class="hljs-keyword">as</span> patches
fig, ax = plt.subplots()
ax.imshow(image)
rectangle = patches.Rectangle((<span class="hljs-number">2350</span>, <span class="hljs-number">1600</span>), <span class="hljs-number">500</span>, <span class="hljs-number">500</span>, linewidth=<span class="hljs-number">2</span>, edgecolor=<span class="hljs-string">&#x27;r&#x27;</span>, facecolor=<span class="hljs-string">&#x27;none&#x27;</span>)
ax.add_patch(rectangle)
ax.axis(<span class="hljs-string">&quot;off&quot;</span>)
plt.show()<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-12t8f97"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/bbox.png" alt="Visualized Bbox"></div> <p data-svelte-h="svelte-1l6eb5k">아래에서 추론 결과를 확인할 수 있습니다.</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 -->fig, ax = plt.subplots()
ax.imshow(image)
ax.imshow(mask, cmap=<span class="hljs-string">&#x27;viridis&#x27;</span>, alpha=<span class="hljs-number">0.4</span>)
ax.axis(<span class="hljs-string">&quot;off&quot;</span>)
plt.show()<!-- HTML_TAG_END --></pre></div> <div class="flex justify-center" data-svelte-h="svelte-1inger9"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/box_inference.png" alt="Visualized Inference"></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/ko/tasks/mask_generation.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_1hrx8 = {
 assets: "/docs/transformers/main/ko",
 base: "/docs/transformers/main/ko",
 env: {}
 };
 const element = document.currentScript.parentElement;
 const data = [null,null];
 Promise.all([
 import("/docs/transformers/main/ko/_app/immutable/entry/start.9aa88961.js"),
 import("/docs/transformers/main/ko/_app/immutable/entry/app.84fb67c3.js")
 ]).then(([kit, app]) => {
 kit.start(app, element, {
 node_ids: [0, 71],
 data,
 form: null,
 error: null
 });
 });
 }
 </script>

Xet Storage Details

Size:
37.9 kB
·
Xet hash:
79a7115d93680de937cf8fcc0a3a59d1d481d42625c994e10f94dfddab9b7f32

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.