Gemaakt met Colaboratory

Author: NielsRogge

SigLIP/Inference_with_(multilingual)_SigLIP,_a_better_CLIP_model.ipynb

@@ -4,7 +4,7 @@
   "metadata": {
     "colab": {
       "provenance": [],
-      "authorship_tag": "ABX9TyNiCJitZw07PsgK/m4d0hNo",
+      "authorship_tag": "ABX9TyML/Sgx5/cEAkxsyW+WQDVc",
       "include_colab_link": true
     },
     "kernelspec": {
@@ -2573,11 +2573,59 @@
         "id": "j40VtILbJmbd"
       }
     },
+    {
+      "cell_type": "markdown",
+      "source": [
+        "## Pipeline API\n",
+        "\n",
+        "If you don't bother about details of pre- and postprocessing, we have the [pipelines](https://huggingface.co/docs/transformers.js/pipelines) which make inference very easy!"
+      ],
+      "metadata": {
+        "id": "lDw5QpSjNidk"
+      }
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "from transformers import pipeline\n",
+        "from PIL import Image\n",
+        "import requests\n",
+        "\n",
+        "# load pipe\n",
+        "image_classifier = pipeline(task=\"zero-shot-image-classification\", model=\"google/siglip-so400m-patch14-384\")\n",
+        "\n",
+        "# load image\n",
+        "url = 'http://images.cocodataset.org/val2017/000000039769.jpg'\n",
+        "image = Image.open(requests.get(url, stream=True).raw)\n",
+        "\n",
+        "# inference\n",
+        "outputs = image_classifier(image, candidate_labels=[\"2 cats\", \"a plane\", \"a remote\"])\n",
+        "outputs = [{\"score\": round(output[\"score\"], 4), \"label\": output[\"label\"] } for output in outputs]\n",
+        "print(outputs)"
+      ],
+      "metadata": {
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        },
+        "id": "jqAmisBiJX_n",
+        "outputId": "e75ad0df-66bb-4bcc-bb11-88b0cc7a422d"
+      },
+      "execution_count": 9,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "[{'score': 0.5089, 'label': '2 cats'}, {'score': 0.0, 'label': 'a remote'}, {'score': 0.0, 'label': 'a plane'}]\n"
+          ]
+        }
+      ]
+    },
     {
       "cell_type": "code",
       "source": [],
       "metadata": {
-        "id": "jqAmisBiJX_n"
+        "id": "Yw4locEcNplK"
       },
       "execution_count": null,
       "outputs": []