116 lines (115 with data), 3.2 kB

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "\n",
    "from wearsed.dataset.WearSEDDataset_vanilla import WearSEDDataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_random_sequence(signals, labels, max_time, seq_length):\n",
    "    start = torch.randint(0, max_time, (1,))\n",
    "    end = start + seq_length\n",
    "    return signals[:, start:end], labels[start:end]\n",
    "\n",
    "def get_batch(signals, labels, batch_size, seq_length):\n",
    "    max_time = signals.shape[1] - seq_length\n",
    "\n",
    "    tries = 0\n",
    "    batch_signals = []\n",
    "    batch_labels = []\n",
    "    for i in range(batch_size):\n",
    "        signals_seq, labels_seq = get_random_sequence(signals, labels, max_time, seq_length)\n",
    "        if i < 0.8*batch_size or tries < 20:  # If there are not enough sequences with events and we haven't tried long enough\n",
    "            if labels_seq.sum() == 0:         # ..check if this is a sequence without events and if so, roll again\n",
    "                signals_seq, labels_seq = get_random_sequence(signals, labels, max_time, seq_length)  # TODO could still be negative sample\n",
    "                batch_signals.append(signals_seq)\n",
    "                batch_labels.append(labels_seq)\n",
    "                tries += 1\n",
    "                continue\n",
    "        batch_signals.append(signals_seq)\n",
    "        batch_labels.append(labels_seq)\n",
    "    return torch.stack(batch_signals), torch.stack(batch_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = WearSEDDataset(mesaid_path='../wearsed/dataset/data_ids/', signals_to_read=['SpO2', 'Pleth'], preprocess=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "42 / 64\n",
      "57 / 64\n",
      "63 / 64\n",
      "37 / 64\n",
      "37 / 64\n",
      "41 / 64\n",
      "62 / 64\n",
      "54 / 64\n",
      "49 / 64\n",
      "48 / 64\n",
      "25 / 64\n",
      "54 / 64\n",
      "52 / 64\n",
      "56 / 64\n",
      "34 / 64\n",
      "56 / 64\n",
      "57 / 64\n",
      "44 / 64\n",
      "37 / 64\n",
      "59 / 64\n"
     ]
    }
   ],
   "source": [
    "for i in range(20):\n",
    "    signals, labels = dataset[i]\n",
    "    _, y = get_batch(signals, labels, batch_size=64, seq_length=600)\n",
    "    print((y.sum(dim=1)>0).sum().item(), '/', y.shape[0])"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "master",
   "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": "3.12.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}