File size: 5,274 Bytes
9665c2c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e30d832c",
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "from pathlib import Path\n",
    "import torch\n",
    "import yaml\n",
    "from torchmetrics import MetricCollection\n",
    "from omegaconf import OmegaConf as OC\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "import numpy as np\n",
    "from pytorch_lightning import seed_everything\n",
    "\n",
    "import maploc\n",
    "from maploc.data import MapillaryDataModule\n",
    "from maploc.data.torch import unbatch_to_device\n",
    "from maploc.module import GenericModule\n",
    "from maploc.models.metrics import Location2DError, AngleError\n",
    "from maploc.evaluation.run import resolve_checkpoint_path\n",
    "from maploc.evaluation.viz import plot_example_single\n",
    "\n",
    "from maploc.models.voting import argmax_xyr, fuse_gps\n",
    "from maploc.osm.viz import Colormap, plot_nodes\n",
    "from maploc.utils.viz_2d import plot_images, features_to_RGB, save_plot, add_text\n",
    "from maploc.utils.viz_localization import likelihood_overlay, plot_pose, plot_dense_rotations, add_circle_inset\n",
    "\n",
    "torch.set_grad_enabled(False);\n",
    "plt.rcParams.update({'figure.max_open_warning': 0})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fc8bd313",
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "conf = OC.load(Path(maploc.__file__).parent / 'conf/data/mapillary.yaml')\n",
    "conf = OC.merge(conf, OC.create(yaml.full_load(\"\"\"\n",
    "data_dir: \"../datasets/MGL_release\"\n",
    "loading:\n",
    "    val: {batch_size: 1, num_workers: 0}\n",
    "    train: ${.val}\n",
    "add_map_mask: true\n",
    "return_gps: true\n",
    "\"\"\")))\n",
    "OC.resolve(conf)\n",
    "dataset = MapillaryDataModule(conf)\n",
    "dataset.prepare_data()\n",
    "dataset.setup()\n",
    "sampler = None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c267ed66",
   "metadata": {},
   "outputs": [],
   "source": [
    "experiment = \"orienternet_mgl.ckpt\"\n",
    "# experiment = \"experiment_name\"  # find the best checkpoint\n",
    "# experiment = \"experiment_name/checkpoint-step=N.ckpt\"  # a given checkpoint\n",
    "path = resolve_checkpoint_path(experiment)\n",
    "print(path)\n",
    "cfg = {'model': {\"num_rotations\": 360, \"apply_map_prior\": True}}\n",
    "model = GenericModule.load_from_checkpoint(\n",
    "    path, strict=True, find_best=not experiment.endswith('.ckpt'), cfg=cfg)\n",
    "model = model.eval().cuda()\n",
    "assert model.cfg.data.resize_image == dataset.cfg.resize_image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "48efb4e3",
   "metadata": {},
   "outputs": [],
   "source": [
    "out_dir = None #Path('outputs_mgl_failures/')\n",
    "if out_dir is not None:\n",
    "    !mkdir -p $out_dir/full\n",
    "\n",
    "seed_everything(25) # best = 25\n",
    "loader = dataset.dataloader(\"val\", shuffle=sampler is None, sampler=sampler)\n",
    "metrics = MetricCollection(model.model.metrics()).to(model.device)\n",
    "metrics[\"xy_gps_error\"] = Location2DError(\"uv_gps\", model.cfg.model.pixel_per_meter)\n",
    "for i, batch in zip(range(10), loader):\n",
    "    pred = data = batch_ = None    \n",
    "    batch_ = model.transfer_batch_to_device(batch, model.device, i)\n",
    "    pred = model(batch_)\n",
    "    pred = {k:v.float() if isinstance(v, torch.HalfTensor) else v for k,v in pred.items()}\n",
    "    pred[\"uv_gps\"] = batch[\"uv_gps\"]\n",
    "    loss = model.model.loss(pred, batch_)\n",
    "    results = metrics(pred, batch_)\n",
    "    results.pop(\"xy_expectation_error\")\n",
    "    for k in list(results):\n",
    "        if \"recall\" in k:\n",
    "            results.pop(k)\n",
    "    print(f'{i} {loss[\"total\"].item():.2f}', {k: round(v.item(), 2) for k, v in results.items()})\n",
    "#     if results[\"xy_max_error\"] < 5:\n",
    "#         continue\n",
    "\n",
    "    pred = unbatch_to_device(pred)\n",
    "    data = unbatch_to_device(batch)\n",
    "    plot_example_single(i, model, pred, data, results, plot_bev=True, out_dir=out_dir, show_gps=True)\n",
    "    \n",
    "    continue\n",
    "    scales_scores = pred['pixel_scales']\n",
    "    log_prob = torch.nn.functional.log_softmax(scales_scores, dim=-1)\n",
    "    scales_exp = torch.sum(log_prob.exp() * torch.arange(scales_scores.shape[-1]), -1)\n",
    "    total_score = torch.logsumexp(scales_scores, -1)\n",
    "    plot_images([log_prob.max(-1).values.exp(), scales_exp, total_score], cmaps='jet')\n",
    "    plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}