-
Notifications
You must be signed in to change notification settings - Fork 110
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Panos Achlioptas
committed
Mar 18, 2018
1 parent
846c628
commit 70177f0
Showing
6 changed files
with
1,183 additions
and
24 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,242 @@ | ||
| { | ||
| "cells": [ | ||
| { | ||
| "cell_type": "markdown", | ||
| "metadata": {}, | ||
| "source": [ | ||
| "# Assuming 2 sets of point-clouds, we will compute the MMD, Coverage and JSD as done in the paper.\n", | ||
| "\n", | ||
| "(To compute these metrics you __don't need__ to have tflearn installed, only the structural: EMD, Chamfer losses and sklearn for the JSD.)" | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "code", | ||
| "execution_count": 1, | ||
| "metadata": { | ||
| "collapsed": false | ||
| }, | ||
| "outputs": [], | ||
| "source": [ | ||
| "import numpy as np\n", | ||
| "import os.path as osp\n", | ||
| "\n", | ||
| "from latent_3d_points.src.evaluation_metrics import minimum_mathing_distance, \\\n", | ||
| "jsd_between_point_cloud_sets, coverage\n", | ||
| "\n", | ||
| "from latent_3d_points.src.in_out import snc_category_to_synth_id,\\\n", | ||
| " load_all_point_clouds_under_folder" | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "markdown", | ||
| "metadata": {}, | ||
| "source": [ | ||
| "Load some point-clouds and make two sets (sample_pcs, ref_pcs) from them. The ref_pcs is considered as the __ground-truth__ data while the sample_pcs corresponds to a set that is matched against it, e.g. comes from a generative model." | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "code", | ||
| "execution_count": 2, | ||
| "metadata": { | ||
| "collapsed": true | ||
| }, | ||
| "outputs": [], | ||
| "source": [ | ||
| "%load_ext autoreload\n", | ||
| "%autoreload 2\n", | ||
| "%matplotlib inline" | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "code", | ||
| "execution_count": 3, | ||
| "metadata": { | ||
| "collapsed": false | ||
| }, | ||
| "outputs": [ | ||
| { | ||
| "name": "stdout", | ||
| "output_type": "stream", | ||
| "text": [ | ||
| "Give me the class name (e.g. \"chair\"): chair\n", | ||
| "6778 pclouds were loaded. They belong in 1 shape-classes.\n" | ||
| ] | ||
| } | ||
| ], | ||
| "source": [ | ||
| "top_in_dir = '../data/shape_net_core_uniform_samples_2048/' # Top-dir of where point-clouds are stored.\n", | ||
| "class_name = raw_input('Give me the class name (e.g. \"chair\"): ').lower()\n", | ||
| "syn_id = snc_category_to_synth_id()[class_name]\n", | ||
| "class_dir = osp.join(top_in_dir , syn_id)\n", | ||
| "all_pc_data = load_all_point_clouds_under_folder(class_dir, n_threads=8, file_ending='.ply', verbose=True)" | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "code", | ||
| "execution_count": 4, | ||
| "metadata": { | ||
| "collapsed": true | ||
| }, | ||
| "outputs": [], | ||
| "source": [ | ||
| "n_ref = 100 # size of ref_pcs.\n", | ||
| "n_sam = 150 # size of sample_pcs.\n", | ||
| "all_ids = np.arange(all_pc_data.num_examples)\n", | ||
| "ref_ids = np.random.choice(all_ids, n_ref, replace=False)\n", | ||
| "sam_ids = np.random.choice(all_ids, n_sam, replace=False)\n", | ||
| "ref_pcs = all_pc_data.point_clouds[ref_ids]\n", | ||
| "sample_pcs = all_pc_data.point_clouds[sam_ids]" | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "markdown", | ||
| "metadata": {}, | ||
| "source": [ | ||
| "Compute the three metrics." | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "code", | ||
| "execution_count": 11, | ||
| "metadata": { | ||
| "collapsed": false | ||
| }, | ||
| "outputs": [], | ||
| "source": [ | ||
| "ae_loss = 'chamfer' # Which distance to use for the matchings.\n", | ||
| "\n", | ||
| "if ae_loss == 'emd':\n", | ||
| " use_EMD = True\n", | ||
| "else:\n", | ||
| " use_EMD = False # Will use Chamfer instead.\n", | ||
| " \n", | ||
| "batch_size = 100 # Find appropriate number that fits in GPU.\n", | ||
| "normalize = True # Matched distances are divided by the number of \n", | ||
| " # points of thepoint-clouds.\n", | ||
| "\n", | ||
| "mmd, matched_dists = minimum_mathing_distance(sample_pcs, ref_pcs, batch_size, normalize=normalize, use_EMD=use_EMD)\n", | ||
| "\n", | ||
| "cov, matched_ids = coverage(sample_pcs, ref_pcs, batch_size, normalize=normalize, use_EMD=use_EMD)\n", | ||
| "\n", | ||
| "jsd = jsd_between_point_cloud_sets(sample_pcs, ref_pcs, resolution=28)" | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "code", | ||
| "execution_count": 12, | ||
| "metadata": { | ||
| "collapsed": false | ||
| }, | ||
| "outputs": [ | ||
| { | ||
| "name": "stdout", | ||
| "output_type": "stream", | ||
| "text": [ | ||
| "0.0714721 0.73 0.0396569736382\n" | ||
| ] | ||
| } | ||
| ], | ||
| "source": [ | ||
| "print mmd, cov, jsd" | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "markdown", | ||
| "metadata": {}, | ||
| "source": [ | ||
| "For a detailed breakdown of the evaluation functions, inspect their docs." | ||
| ] | ||
| }, | ||
| { | ||
| "cell_type": "code", | ||
| "execution_count": 24, | ||
| "metadata": { | ||
| "collapsed": false | ||
| }, | ||
| "outputs": [ | ||
| { | ||
| "name": "stdout", | ||
| "output_type": "stream", | ||
| "text": [ | ||
| "Computes the Coverage between two sets of point-clouds.\n", | ||
| "\n", | ||
| " Args:\n", | ||
| " sample_pcs (numpy array SxKx3): the S point-clouds, each of K points that will be matched\n", | ||
| " and compared to a set of \"reference\" point-clouds.\n", | ||
| " ref_pcs (numpy array RxKx3): the R point-clouds, each of K points that constitute the\n", | ||
| " set of \"reference\" point-clouds.\n", | ||
| " batch_size (int): specifies how large will the batches be that the compute will use to\n", | ||
| " make the comparisons of the sample-vs-ref point-clouds.\n", | ||
| " normalize (boolean): When the matching is based on Chamfer (default behavior), if True,\n", | ||
| " the Chamfer is computed as the average of the matched point-wise squared euclidean\n", | ||
| " distances. Alternatively, is their sum.\n", | ||
| " use_sqrt (boolean): When the matching is based on Chamfer (default behavior), if True,\n", | ||
| " the Chamfer is computed based on the (not-squared) euclidean distances of the matched\n", | ||
| " point-wise euclidean distances.\n", | ||
| " sess (tf.Session): If None, it will make a new Session for this.\n", | ||
| " use_EMD (boolean): If true, the matchings are based on the EMD.\n", | ||
| " ret_dist (boolean): If true, it will also return the distances between each sample_pcs and\n", | ||
| " it's matched ground-truth.\n", | ||
| " Returns: the coverage score (int),\n", | ||
| " the indices of the ref_pcs that are matched with each sample_pc\n", | ||
| " and optionally the matched distances of the samples_pcs.\n", | ||
| " \n", | ||
| "Computes the MMD between two sets of point-clouds.\n", | ||
| "\n", | ||
| " Args:\n", | ||
| " sample_pcs (numpy array SxKx3): the S point-clouds, each of K points that will be matched and\n", | ||
| " compared to a set of \"reference\" point-clouds.\n", | ||
| " ref_pcs (numpy array RxKx3): the R point-clouds, each of K points that constitute the set of\n", | ||
| " \"reference\" point-clouds.\n", | ||
| " batch_size (int): specifies how large will the batches be that the compute will use to make\n", | ||
| " the comparisons of the sample-vs-ref point-clouds.\n", | ||
| " normalize (boolean): When the matching is based on Chamfer (default behavior), if True, the\n", | ||
| " Chamfer is computed as the average of the matched point-wise squared euclidean distances.\n", | ||
| " Alternatively, is their sum.\n", | ||
| " use_sqrt: (boolean): When the matching is based on Chamfer (default behavior), if True, the\n", | ||
| " Chamfer is computed based on the (not-squared) euclidean distances of the matched point-wise\n", | ||
| " euclidean distances.\n", | ||
| " sess (tf.Session, default None): if None, it will make a new Session for this.\n", | ||
| " use_EMD (boolean: If true, the matchings are based on the EMD.\n", | ||
| "\n", | ||
| " Returns:\n", | ||
| " A tuple containing the MMD and all the matched distances of which the MMD is their mean.\n", | ||
| " \n", | ||
| " JSD between two sets of point-clouds, as introduced in the paper ```Learning Representations And Generative Models For 3D Point Clouds```. \n", | ||
| " Args:\n", | ||
| " sample_pcs: (np.ndarray S1xR2x3) S1 point-clouds, each of R1 points.\n", | ||
| " ref_pcs: (np.ndarray S2xR2x3) S2 point-clouds, each of R2 points.\n", | ||
| " resolution: (int) grid-resolution. Affects granularity of measurements.\n", | ||
| " \n" | ||
| ] | ||
| } | ||
| ], | ||
| "source": [ | ||
| "print coverage.__doc__\n", | ||
| "print minimum_mathing_distance.__doc__\n", | ||
| "print jsd_between_point_cloud_sets.__doc__" | ||
| ] | ||
| } | ||
| ], | ||
| "metadata": { | ||
| "kernelspec": { | ||
| "display_name": "TensorFlow1", | ||
| "language": "python", | ||
| "name": "tf1" | ||
| }, | ||
| "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": 2 | ||
| } |
Oops, something went wrong.