Mercurial Hosting > traffic-intelligence
changeset 1219:8a626226793e
update where optimization uses either nomad-parameter file depending on optimizing 1 or 2 steps
| author | Nicolas Saunier <nicolas.saunier@polymtl.ca> |
|---|---|
| date | Mon, 19 Jun 2023 17:09:56 -0400 |
| parents | 1f0b1fc172f8 |
| children | 2dc9899b73ae |
| files | scripts/dltrack.py scripts/nomad/nomad-parameters-4.txt scripts/nomad/nomad-parameters.txt scripts/nomad/optimize-with-nomad.py scripts/nomad/site-parameters-optimization.py trafficintelligence/moving.py |
| diffstat | 6 files changed, 91 insertions(+), 63 deletions(-) [+] |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/scripts/dltrack.py Mon Jun 19 17:09:56 2023 -0400 @@ -0,0 +1,27 @@ +#! /usr/bin/env python3 +# from https://docs.ultralytics.com/modes/track/ +import sys, argparse + +from trafficintelligence.moving import cocoUserTypeNames +from ultralytics import YOLO + +parser = argparse.ArgumentParser(description='The program tracks objects following the ultralytics yolo executable.')#, epilog = 'Either the configuration filename or the other parameters (at least video and database filenames) need to be provided.') +parser.add_argument('-i', dest = 'videoFilename', help = 'name of the video file (overrides the configuration file)') +# detect model +# tracker model +parser.add_argument('--display', dest = 'display', help = 'show the results (careful with long videos, risk of running out of memory)', action = 'store_true') +args = parser.parse_args() + +# Load a model +model = YOLO('/home/nicolas/Research/Data/classification-models/yolov8x.pt ') # seg yolov8x-seg.pt +# seg could be used on cropped image... if can be loaded and kept in memory +# model = YOLO('/home/nicolas/Research/Data/classification-models/yolo_nas_l.pt ') # AttributeError: 'YoloNAS_L' object has no attribute 'get' + +# Track with the model +if args.display: + results = model.track(source=args.videoFilename, tracker="/home/nicolas/Research/Data/classification-models/bytetrack.yaml", classes=list(cocoUserTypeNames.keys()), show=True) # , save_txt=True +else: + results = model.track(source=args.videoFilename, tracker="/home/nicolas/Research/Data/classification-models/bytetrack.yaml", classes=list(cocoUserTypeNames.keys()), stream=True) + for result in results: + for box in result.boxes: + print(box.xyxy)
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/scripts/nomad/nomad-parameters-4.txt Mon Jun 19 17:09:56 2023 -0400 @@ -0,0 +1,30 @@ +DIMENSION 4 # number of variables + +BB_EXE "$python3 site-parameters-optimization.py" # 'site-parameters-optimization.py' is the blackbox program +BB_OUTPUT_TYPE OBJ # object will be minimized + +X0 initial-parameters-4.txt # starting point + +LOWER_BOUND ( 2 0.1 0.1 1.0 ) # all variables' lower bounds +UPPER_BOUND ( 100 10 5 15 ) # all variables' upper bounds + +MAX_BB_EVAL 500 # the algorithm terminates when + # n black-box evaluations have + # been made + +# TMP_DIR /tmp # indicates a directory where + # temporary files are put + # (increases performance by ~100% + # if you're working on a network + # account and if TMP_DIR is on a + # local disk) + +DISPLAY_DEGREE 2 + +# DISPLAY_ALL_EVAL yes + +DISPLAY_STATS BBE ( SOL ) OBJ # Display the number of evaluation (BBE), + # the current solution ( SOL ) and the objective + +# STATS_FILE test.txt BBE ( SOL ) OBJ +HISTORY_FILE history.txt
--- a/scripts/nomad/nomad-parameters.txt Sat Jun 17 08:48:22 2023 -0400 +++ b/scripts/nomad/nomad-parameters.txt Mon Jun 19 17:09:56 2023 -0400 @@ -5,7 +5,7 @@ X0 initial-parameters.txt # starting point -LOWER_BOUND ( 0 0 1 0.01 2 0.1 0.1 1.01 ) # all variables' lower bounds +LOWER_BOUND ( 0 0 1 0.01 2 0.1 0.1 1.0) # all variables' lower bounds UPPER_BOUND ( 1 10 10 0.3 100 10 5 15 ) # all variables' upper bounds MAX_BB_EVAL 500 # the algorithm terminates when
--- a/scripts/nomad/optimize-with-nomad.py Sat Jun 17 08:48:22 2023 -0400 +++ /dev/null Thu Jan 01 00:00:00 1970 +0000 @@ -1,29 +0,0 @@ -import os -import argparse - -parser = argparse.ArgumentParser(description='The program is used to select the type of tracking to run with ' - 'tracking-mota.py with NOMAD', - epilog='''NOMAD - A blackbox optimization software: - C. Audet, S. Le Digabel, C. Tribes and V. Rochon Montplaisir. The NOMAD project. - Software available at https://www.gerad.ca/nomad. - - S. Le Digabel. Algorithm 909: NOMAD: Nonlinear Optimization with the MADS algorithm. - ACM Transactions on Mathematical Software, 37(4):44:1–44:15, 2011.''', - formatter_class=argparse.RawDescriptionHelpFormatter) - -parser.add_argument('-t', dest='intersections', nargs = '*', type = str, help='name of the intersection for which ' - 'the optimization is meant to be ran', - required=True) -parser.add_argument('--optimize-grouping-only', dest='optimizeGroupingOnly', - help='optimize only the grouping parameters and not the ones associated with feature tracking', - action='store_true') - -args = parser.parse_args() - -with open('arguments.txt', 'w') as f: - f.write(str(args.intersections)) - f.write("\n") - f.write(str(args.optimizeGroupingOnly)) - -os.system('nomad nomad-parameters.txt initial-parameters.txt') -
--- a/scripts/nomad/site-parameters-optimization.py Sat Jun 17 08:48:22 2023 -0400 +++ b/scripts/nomad/site-parameters-optimization.py Mon Jun 19 17:09:56 2023 -0400 @@ -7,7 +7,7 @@ import numpy as np -def loadParametersStartProcess(filename): +def loadParametersStartProcess(filename, intersections): # load initial parameters from x.txt f = open(filename, 'r+') l = f.readline() @@ -18,13 +18,12 @@ para = paraValueList(x) # run process including trackingfeature, groupfeature, load groundtruth, compute mota - print(process(para, intersections, optimizeGroupingOnly)) + print(process(para, intersections)) def paraValueList(x): #create para-value list #list of the 8 parameters and their values - pn = 8 - p = pn*[None] + p = 8*[None] p[0] = '--feature-quality' #]0.-0.4] p[1] = '--min-feature-distanceklt' #]0.-6] p[2] = '--window-size' #[1-10]integer @@ -35,12 +34,16 @@ p[7] = '--min-nfeatures-group' #[2-4] para = [] - for n in range(pn): - para = para + [p[n],x[n]] + if len(x) == 4: + for n in range(4): + para = para + [p[4+n],x[n]] + else: + for n in range(len(x)): + para = para + [p[n],x[n]] return para -def process(para, intersections, optimizeGroupingOnly): +def process(para, intersections): Mota = [] gtDatabaseaAbsPaths = [] configFileAbsPaths = [] @@ -58,13 +61,15 @@ os.chdir(cwd) for gtDatabaseAbsPath, configFileAbsPath in zip(gtDatabaseaAbsPaths, configFileAbsPaths): gtDatabaseBasename = gtDatabaseAbsPath[:-10] - videoFilename = gtDatabaseBasename + ".MP4" + videoFilename = gtDatabaseBasename + ".avi" # careful, it may be necessary to check video type / extension + if not os.path.exists(videoFilename): + print('Video file {} does not exist'.format(videoFilename)) databaseFilename = gtDatabaseBasename + ".sqlite" gtDatabaseDirname = os.path.dirname(gtDatabaseAbsPath) homographyFilename = gtDatabaseDirname + "/homography.txt" maskFilename = gtDatabaseDirname + "/mask.png" # Skip feature tracking if the user specified to optimize only grouping parameters - if not optimizeGroupingOnly: + if len(para) > 8: # Track features trackingFeature(para, configFileAbsPath, videoFilename, databaseFilename, homographyFilename, maskFilename) # Group features @@ -91,7 +96,7 @@ # trackingfeature command parameters tf = ['feature-based-tracking', config, '--tf', '--video-filename', video, '--database-filename', db, '--homography-filename', homo, '--mask-filename', mask] # run in command line and print directly - subprocess.check_output(tf + para[0:10]) + subprocess.check_output(tf + para) def groupFeature(para, config, video, db, homo, mask): #remove previous grouping @@ -99,24 +104,19 @@ #groupfeature command parameters gf = ['feature-based-tracking', config, '--gf', '--video-filename', video, '--database-filename', db, '--homography-filename', homo, '--mask-filename', mask] #run in command line and print directly - subprocess.check_output(gf + para[8:16]) # ['--min-feature-time', 'x', '--mm-connection-distance', 'x', '--mm-segmentation-distance', 'x', '--min-nfeatures-group', 'x'] - -def computeMota(annotations, objects, Mota): - matchingDistance = 5 - firstInstant = 0 - lastInstant = 50000 - return moving.computeClearMOT(annotations, objects, matchingDistance, firstInstant, lastInstant)[1] + subprocess.check_output(gf + para) # ['--min-feature-time', 'x', '--mm-connection-distance', 'x', '--mm-segmentation-distance', 'x', '--min-nfeatures-group', 'x'] if __name__ == "__main__": # Load args that were given with select-arguments.py - with open('arguments.txt', 'r') as f: - args = f.read().split('\n') - intersections = args[0] - optimizeGroupingOnly = args[1] - # Convert string representation of list into list - intersections = eval(intersections) + # with open('arguments.txt', 'r') as f: + # args = f.read().split('\n') + # intersections = args[0] + # optimizeGroupingOnly = eval(args[1]) + # intersections = eval(intersections) + + # Just write the intersections to optimize here + intersections = ['../12-laurier'] - loadParametersStartProcess(sys.argv[1]) + loadParametersStartProcess(sys.argv[1], intersections) sys.exit(0) -
--- a/trafficintelligence/moving.py Sat Jun 17 08:48:22 2023 -0400 +++ b/trafficintelligence/moving.py Mon Jun 19 17:09:56 2023 -0400 @@ -1363,14 +1363,14 @@ 'truck', 'automated'] -coco2UserTypes = {1: 2, 2: 4, 3: 1, 6: 5, 8: 6} -cocoUserTypeNames = {1: person, - 2: bicycle, - 3: car, - 4: motorcycle, - 6: bus, - 7: train, - 8: truck} +coco2UserTypes = {0: 2, 1: 4, 2: 1, 5: 5, 7: 6} +cocoUserTypeNames = {0: 'person', + 1: 'bicycle', + 2: 'car', + 3: 'motorcycle', + 5: 'bus', + 6: 'train', + 7: 'truck'} userType2Num = utils.inverseEnumeration(userTypeNames)