Verify the behavior of `smp.losses.JaccardLoss`.

jaccard_test.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "from segmentation_models_pytorch.losses import JaccardLoss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = torch.rand(1, 4, 256, 256)\n",
    "y_pred[:,1,:,:] = 0\n",
    "y_pred = nn.functional.softmax(y_pred, dim=1)\n",
    "\n",
    "y_true = torch.randint(0, 4, size=(1, 256, 256))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(0.8589)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loss = JaccardLoss(mode=\"multiclass\", classes=None, log_loss=False, from_logits=False)\n",
    "loss(y_pred, y_true)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(0.8589)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loss = JaccardLoss(mode=\"multiclass\", classes=4, log_loss=False, from_logits=False)\n",
    "loss(y_pred, y_true)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(0.8490)\n",
      "tensor(0.8893)\n",
      "tensor(0.8479)\n",
      "tensor(0.8493)\n"
     ]
    }
   ],
   "source": [
    "for i in range(4):\n",
    "    loss = JaccardLoss(mode=\"multiclass\", classes=[i], log_loss=False, from_logits=False)\n",
    "    print(loss(y_pred, y_true))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(0.8589)\n"
     ]
    }
   ],
   "source": [
    "loss = JaccardLoss(mode=\"multiclass\", classes=[0,1,2,3], log_loss=False, from_logits=False)\n",
    "print(loss(y_pred, y_true))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(0.8487)\n"
     ]
    }
   ],
   "source": [
    "loss = JaccardLoss(mode=\"multiclass\", classes=[0,2,3], log_loss=False, from_logits=False)\n",
    "print(loss(y_pred, y_true))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "geo",
   "language": "python",
   "name": "geo"
  },
  "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
}