Mercurial Hosting > traffic-intelligence
view scripts/compute-clearmot.py @ 1117:bb58506c1bfd
minor
author | Nicolas Saunier <nicolas.saunier@polymtl.ca> |
---|---|
date | Tue, 25 Jun 2019 00:38:56 -0400 |
parents | cc5cb04b04b0 |
children | 7eb972942f22 |
line wrap: on
line source
#! /usr/bin/env python3 import sys, argparse from numpy import loadtxt from numpy.linalg import inv from trafficintelligence import moving, storage, cvutils # TODO: need to trim objects to same mask ? parser = argparse.ArgumentParser(description='The program computes the CLEAR MOT metrics between ground truth and tracker output (in Polytrack format)', epilog='''CLEAR MOT metrics information: Keni, Bernardin, and Stiefelhagen Rainer. "Evaluating multiple object tracking performance: the CLEAR MOT metrics." EURASIP Journal on Image and Video Processing 2008 (2008) Polytrack format: JP. Jodoin\'s MSc thesis (in french) see examples on http://www.jpjodoin.com/urbantracker/dataset.html''', formatter_class=argparse.RawDescriptionHelpFormatter) parser.add_argument('-d', dest = 'trackerDatabaseFilename', help = 'name of the Sqlite database containing the tracker output', required = True) parser.add_argument('-g', dest = 'groundTruthDatabaseFilename', help = 'name of the Sqlite database containing the ground truth', required = True) parser.add_argument('-o', dest = 'homographyFilename', help = 'name of the filename for the homography (if tracking was done using the homography)') parser.add_argument('-m', dest = 'matchingDistance', help = 'matching distance between tracker and ground truth trajectories', required = True, type = float) parser.add_argument('--mask', dest = 'maskFilename', help = 'filename of the mask file used to define the where objects were tracked') parser.add_argument('-f', dest = 'firstInstant', help = 'first instant for measurement', required = True, type = int) parser.add_argument('-l', dest = 'lastInstant', help = 'last instant for measurement', required = True, type = int) parser.add_argument('--offset', dest = 'nFramesOffsetAnnotations', help = 'number of frames to offset the ground truth annotations', type = int) parser.add_argument('--displayOffset', dest = 'nFramesOffsetDisplay', help = 'number of frames to offset annotations and objects for display', type = int) parser.add_argument('--display', dest = 'display', help = 'display the ground truth to object matches (graphically)', action = 'store_true') parser.add_argument('--undistort', dest = 'undistort', help = 'undistort the video (because features have been extracted that way)', action = 'store_true') parser.add_argument('--intrinsic', dest = 'intrinsicCameraMatrixFilename', help = 'name of the intrinsic camera file') parser.add_argument('--distortion-coefficients', dest = 'distortionCoefficients', help = 'distortion coefficients', nargs = '*', type = float) parser.add_argument('--undistorted-multiplication', dest = 'undistortedImageMultiplication', help = 'undistorted image multiplication', type = float) parser.add_argument('-i', dest = 'videoFilename', help = 'name of the video file (for display)') parser.add_argument('--csv', dest = 'csvOutput', help = 'output comma-separated metrics', action = 'store_true') args = parser.parse_args() if args.homographyFilename is not None: invHomography = inv(loadtxt(args.homographyFilename)) else: invHomography = None objects = storage.loadTrajectoriesFromSqlite(args.trackerDatabaseFilename, 'object') if args.maskFilename is not None: maskObjects = [] from matplotlib.pyplot import imread mask = imread(args.maskFilename) if len(mask) > 1: mask = mask[:,:,0] for obj in objects: maskObjects += obj.getObjectsInMask(mask, invHomography, 10) # TODO add option to keep object if at least one feature in mask objects = maskObjects annotations = storage.loadBBMovingObjectsFromSqlite(args.groundTruthDatabaseFilename) for a in annotations: a.computeCentroidTrajectory(invHomography) if args.nFramesOffsetAnnotations is not None: for a in annotations: a.shiftTimeInterval(args.nFramesOffsetAnnotations) motp, mota, mt, mme, fpt, gt, gtMatches, toMatches = moving.computeClearMOT(annotations, objects, args.matchingDistance, args.firstInstant, args.lastInstant, args.display) if args.csvOutput: print('{},{},{},{},{}'.format(motp, mota, mt, mme, fpt)) else: print 'MOTP: {}'.format(motp) print 'MOTA: {}'.format(mota) print 'Number of missed objects.frames: {}'.format(mt) print 'Number of mismatches: {}'.format(mme) print 'Number of false alarms.frames: {}'.format(fpt) def shiftMatches(matches, offset): shifted = {} for k in matches: shifted[k] = {t+offset:v for t, v in matches[k].items()} return shifted if args.display: if args.undistort and args.intrinsicCameraMatrixFilename is not None: intrinsicCameraMatrix = loadtxt(args.intrinsicCameraMatrixFilename) else: intrinsicCameraMatrix = None firstInstant = args.firstInstant lastInstant = args.lastInstant cvutils.displayTrajectories(args.videoFilename, objects, {}, invHomography, firstInstant, lastInstant, annotations = annotations, undistort = args.undistort, intrinsicCameraMatrix = intrinsicCameraMatrix, distortionCoefficients = args.distortionCoefficients, undistortedImageMultiplication = args.undistortedImageMultiplication, gtMatches = gtMatches, toMatches = toMatches) #print('Ground truth matches') #print(gtMatches) #print('Object matches') #rint toMatches