{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "1d923e4f84643d2b",
   "metadata": {},
   "source": "# Test time augmentation"
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "620ed78a77d63983",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-08-27T10:59:02.570959Z",
     "start_time": "2024-08-27T10:59:02.527272Z"
    }
   },
   "outputs": [],
   "source": [
    "# inspecting a sample tta configuration\n",
    "import mml.configs\n",
    "from pathlib import Path\n",
    "from omegaconf import OmegaConf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "c905c0e1e7a6aaf4",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2024-08-27T11:00:53.129520Z",
     "start_time": "2024-08-27T11:00:53.118971Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mode: mean\n",
      "variations:\n",
      "  identity:\n",
      "  - name: RandomHorizontalFlip\n",
      "    p: 0.0\n",
      "  rot90:\n",
      "  - name: RandomRotation\n",
      "    degrees:\n",
      "    - 90\n",
      "    - 90\n",
      "    p: 1.0\n",
      "  rot270:\n",
      "  - name: RandomRotation\n",
      "    degrees:\n",
      "    - 270\n",
      "    - 270\n",
      "    p: 1.0\n",
      "  hflip:\n",
      "  - name: RandomHorizontalFlip\n",
      "    p: 1.0\n",
      "\n"
     ]
    }
   ],
   "source": [
    "path = Path(mml.configs.__file__).parent / \"tta\" / \"rotate.yaml\"\n",
    "cfg = OmegaConf.load(path)\n",
    "print(OmegaConf.to_yaml(cfg))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ee3966b85422869f",
   "metadata": {},
   "source": "`mode` determines the merging strategy, currently only `mean` is supported but other strategies might be implemented later (majority vote, etc.). `variations` lists all [kornia augmentations](https://kornia.readthedocs.io/en/latest/augmentation.module.html) that shall be performed - ideally they are deterministic. `mml` will apply each transform to each image and merge the predictions of the model according to the `mode`. Note that `tta` is only active during testing and predicting with models - not during training nor validation."
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1c0c31de9f762235",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}