{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "from PIL import Image\n",
    "\n",
    "import torch\n",
    "from transformers import (\n",
    "    AutoModelForImageClassification,\n",
    "    AutoImageProcessor,\n",
    ")\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "MODEL_NAME = \"p1atdev/siglip-tagger-test-2\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = AutoModelForImageClassification.from_pretrained(\n",
    "    MODEL_NAME, torch_dtype=torch.bfloat16, trust_remote_code=True\n",
    ")\n",
    "model.eval()\n",
    "processor = AutoImageProcessor.from_pretrained(MODEL_NAME)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "image = Image.open(\"sample.jpg\")\n",
    "inputs = processor(image, return_tensors=\"pt\").to(model.device, model.dtype)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "logits = model(**inputs).logits.detach().cpu().float()[0]\n",
    "logits = np.clip(logits, 0.0, 1.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "results = {\n",
    "    model.config.id2label[i]: logit for i, logit in enumerate(logits) if logit > 0\n",
    "}\n",
    "results = sorted(results.items(), key=lambda x: x[1], reverse=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1girl: 100.00%\n",
      "outdoors: 100.00%\n",
      "sky: 100.00%\n",
      "solo: 100.00%\n",
      "school uniform: 96.88%\n",
      "skirt: 92.97%\n",
      "day: 89.06%\n",
      "cloud: 85.94%\n",
      "scenery: 79.69%\n",
      "pleated skirt: 72.27%\n",
      "black hair: 66.80%\n",
      "standing: 65.62%\n",
      "sailor collar: 59.38%\n",
      "sitting: 57.81%\n",
      "long sleeves: 53.52%\n",
      "serafuku: 53.12%\n",
      "holding: 52.34%\n",
      "tree: 47.46%\n",
      "dress: 46.48%\n",
      "shoes: 43.55%\n",
      "building: 42.77%\n",
      "neckerchief: 40.82%\n",
      "short hair: 38.09%\n",
      "water: 38.09%\n",
      "cloudy sky: 37.30%\n",
      "looking at viewer: 32.23%\n",
      "long hair: 32.03%\n",
      "brown eyes: 31.45%\n",
      "plant: 31.05%\n",
      "bag: 29.30%\n",
      "railing: 29.10%\n",
      "sunlight: 28.12%\n",
      "from side: 27.73%\n",
      "window: 27.54%\n",
      "brown hair: 26.37%\n",
      "white shirt: 25.78%\n",
      "shirt: 25.39%\n",
      "blue sky: 23.93%\n",
      "hairclip: 23.44%\n",
      "blunt bangs: 21.58%\n",
      "picture frame: 19.34%\n",
      "hand up: 18.26%\n",
      "black skirt: 17.87%\n",
      "smile: 17.87%\n",
      "from behind: 13.57%\n",
      "cowboy shot: 10.99%\n",
      "indoors: 10.74%\n",
      "curtains: 10.25%\n",
      "facing away: 9.23%\n",
      "white socks: 6.08%\n",
      "bottle: 6.01%\n",
      "mountain: 5.66%\n",
      "blue skirt: 5.13%\n",
      "drinking straw: 3.37%\n",
      "kneehighs: 1.71%\n"
     ]
    }
   ],
   "source": [
    "for tag, score in results:\n",
    "    print(f\"{tag}: {score*100:.2f}%\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "py310",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}