Add fast-fg-est refinement as a postprocessing before putting alpha mat.

tutorials/BiRefNet_inference.ipynb

@@ -134,7 +134,6 @@
       "source": [
         "import os\n",
         "from glob import glob\n",
-        "import numpy as np\n",
         "from image_proc import refine_foreground\n",
         "\n",
         "src_dir = '../images_todo'\n",
@@ -168,41 +167,6 @@
         "display(image_masked.resize(scaled_size))"
       ]
     },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "id": "6bWgo0JS-eqF"
-      },
-      "outputs": [],
-      "source": [
-        "# import requests\n",
-        "# from io import BytesIO\n",
-        "\n",
-        "\n",
-        "# image_url = \"https://qph.cf2.quoracdn.net/main-qimg-d89362d538d350a4e4218a366e0d0425-pjlq\"\n",
-        "# response = requests.get(image_url)\n",
-        "# image_data = BytesIO(response.content)\n",
-        "# image = Image.open(image_data)\n",
-        "# input_images = transform_image(image).unsqueeze(0).to(device)\n",
-        "\n",
-        "# # Prediction\n",
-        "# with torch.no_grad():\n",
-        "#     preds = birefnet(input_images)[-1].sigmoid().cpu()\n",
-        "# pred = preds[0].squeeze()\n",
-        "\n",
-        "# # Show Results\n",
-        "# pred_pil = transforms.ToPILImage()(pred)\n",
-        "# # Scale proportionally with max length to 1024 for faster showing\n",
-        "# scale_ratio = 1024 / max(image.size)\n",
-        "# scaled_size = (int(image.size[0] * scale_ratio), int(image.size[1] * scale_ratio))\n",
-        "# image_masked = refine_foreground(image, pred_pil)\n",
-        "# image_masked.putalpha(pred_pil.resize(image.size))\n",
-        "# display(image.resize(scaled_size))\n",
-        "# display(pred_pil.resize(scaled_size))\n",
-        "# display(image_masked.resize(scaled_size))"
-      ]
-    },
     {
       "cell_type": "code",
       "execution_count": null,