Ejemplo n.º 1
0
def dp_dist_fps(config, parser): #@UnusedVariable
    """ Computation speed statistics for distances. """
    parser.add_option("-d", "--distances", default='*',
                      help="Comma-separated list of distances. Can use *.")
    
    parser.add_option("-t", "--testcases", default='*',
                       help="Comma-separated list of algorithms. Can use *.")

    parser.add_option("-r", "--repeat", default=1, type='int',
                       help="Repeat many times.")
 
    options = parser.parse_options()
    
    distances = config.distances.expand_names(options.distances)
    testcases = config.testcases.expand_names(options.testcases)

    logger.info('Using distances: %s' % distances)
    logger.info('Using testcases: %s' % testcases)
    
    testcases = map(config.testcases.instance, testcases)
    
    for id_distance in distances:    
        dist = config.distances.instance(id_distance)
        
        fps = InAWhile()
        for _ in range(options.repeat):
            for tc in testcases: 
                fps.its_time()
                result = dist.distance(tc.y0, tc.y1)
                print('%s / %s: %s (%s)' % (id_distance, tc.id_tc, result, tc.true_plan))
        print('frames per second: %s' % fps.fps())
Ejemplo n.º 2
0
def measure_fps():
    n = 100
    for id_data, id_algo in itertools.product(data, algos):
        v = data[id_data]
        algo = algos[id_algo]
    
        count = InAWhile()
        for _ in xrange(n):
            algo(v)
            count.its_time()
        
        fps = count.fps()
        print('%s-%15s: %s' % (id_data, id_algo, fps))
def diffeo_quick1(id_dd, dd):  # @UnusedVariable
    M, N = dd.shape[0], dd.shape[1]
    template = np.random.rand(M, N)

    fda = FastDiffeoApply(dd)

    functions = dict(regular=lambda: diffeo_apply(dd, template),
                     fast=lambda: fda(template))

    repeat = 100
    for name, function in functions.items():
        fps = InAWhile()
        for _ in range(repeat):
            fps.its_time()
            _ = function()
        print('%s: %s fps' % (name, fps.fps()))
def diffeo_quick1(id_dd, dd):  # @UnusedVariable
    M, N = dd.shape[0], dd.shape[1]
    template = np.random.rand(M, N)

    fda = FastDiffeoApply(dd)
    
    functions = dict(regular=lambda: diffeo_apply(dd, template),
                     fast=lambda: fda(template))

    repeat = 100
    for name, function in functions.items():
        fps = InAWhile()
        for _ in range(repeat):
            fps.its_time()
            _ = function()
        print('%s: %s fps' % (name, fps.fps()))
Ejemplo n.º 5
0
def dp_dist_fps(config, parser):  #@UnusedVariable
    """ Computation speed statistics for distances. """
    parser.add_option("-d",
                      "--distances",
                      default='*',
                      help="Comma-separated list of distances. Can use *.")

    parser.add_option("-t",
                      "--testcases",
                      default='*',
                      help="Comma-separated list of algorithms. Can use *.")

    parser.add_option("-r",
                      "--repeat",
                      default=1,
                      type='int',
                      help="Repeat many times.")

    options = parser.parse_options()

    distances = config.distances.expand_names(options.distances)
    testcases = config.testcases.expand_names(options.testcases)

    logger.info('Using distances: %s' % distances)
    logger.info('Using testcases: %s' % testcases)

    testcases = map(config.testcases.instance, testcases)

    for id_distance in distances:
        dist = config.distances.instance(id_distance)

        fps = InAWhile()
        for _ in range(options.repeat):
            for tc in testcases:
                fps.its_time()
                result = dist.distance(tc.y0, tc.y1)
                print('%s / %s: %s (%s)' %
                      (id_distance, tc.id_tc, result, tc.true_plan))
        print('frames per second: %s' % fps.fps())
Ejemplo n.º 6
0
def diffeo_learner_main():
    usage = "usage: %prog   -p outputpath -n ddsname -s [W,H] -a [rx,ry] bag1 bag2 .."
    parser = OptionParser(usage=usage, version="%prog 1.0") 
    parser.add_option("-p", "--path", default=None,
                      help="Path to output files")
    parser.add_option("-n", "--name", default='camdds',
                      help="Output dds system name")
    
    parser.add_option("-s", "--size", default=None, # default='[160,120]',
                      help="Image size WxH")
    parser.add_option("-a", "--area", default=None, #default='[6,6]',
                      help="Size of search area")
    
    parser.add_option("-r", "--ratio", default=0.1, type='float',
                      help="Search area ratio.")
    
    parser.add_option("-l", "--learner", default='No',
                      help="Reuse learner")
    parser.add_option("-o", "--outlearner", default='No',
                      help="Save the learner")
    parser.add_option("-r", "--refine", default=0,
                      help="Refine the learning by running through the logs r times")
    options, args = parser.parse_args()
    
    if options.path is None:
        dirname = os.path.dirname(options.input)
        logger.info('No path given; using %s' % dirname)
    else:
        dirname = options.path
         
    name = options.name 

    if options.size or options.area:
        msg = 'Using deprecated options. Start using --ratio.'
        raise UserError(msg)
        


    learner = options.learner
    outlearner = options.outlearner

    if learner == 'No':
        ratios = (options.ratio, options.ratio)
        
        learn = DiffeoLearner(True, {'max_displ': ratios, 'inference_method': 'sim'})
    else:
        logger.info('Loading diffeomorphism estimators')
        if learner[:4] == 'http':
            logger.info('from online source: %s' % learner)
            ofile = urllib.urlopen(learner)
        else:
            logger.info('from local source: %s' % learner)
            ofile = open(learner)
        learn = pickle.load(ofile)
        logger.debug('loaded')
    
    in_a_while = InAWhile(30)
    
    
    refine = int(options.refine)
    for _ in range(refine):
        for bagfile in args:
            i = 0
            for y0, u, y1 in read_bag(bagfile):    
                logger.info('Iteration number %d' % i)
                learn.update(y0, u, y1)
                i += 1
                
                if in_a_while.its_time():
                    display_current_results(learn, name, dirname, i)
                    in_a_while.reset()
                
                if i > 15:
                    break
            
        if outlearner == 'No':
            pass
        else:
            logger.info('Saving learning agent to %s' % outlearner)
            pickle.dump(learn, open(outlearner, 'wb'))
    
        
        logger.info('Commands: %s' % learn.command_list)
        learn.summarize(prefix=name)
        
    learn.analyze(prefix=name, folder='out/diffeo-analysis/')
    
    learn.diffeo_dump(dirname, name)
Ejemplo n.º 7
0
def diffeo_learner_main():
    usage = "usage: %prog   -p outputpath -n ddsname -s [W,H] -a [rx,ry] bag1 bag2 .."
    parser = OptionParser(usage=usage, version="%prog 1.0")
    parser.add_option("-p",
                      "--path",
                      default=None,
                      help="Path to output files")
    parser.add_option("-n",
                      "--name",
                      default='camdds',
                      help="Output dds system name")

    parser.add_option(
        "-s",
        "--size",
        default=None,  # default='[160,120]',
        help="Image size WxH")
    parser.add_option(
        "-a",
        "--area",
        default=None,  #default='[6,6]',
        help="Size of search area")

    parser.add_option("-r",
                      "--ratio",
                      default=0.1,
                      type='float',
                      help="Search area ratio.")

    parser.add_option("-l", "--learner", default='No', help="Reuse learner")
    parser.add_option("-o",
                      "--outlearner",
                      default='No',
                      help="Save the learner")
    parser.add_option(
        "-r",
        "--refine",
        default=0,
        help="Refine the learning by running through the logs r times")
    options, args = parser.parse_args()

    if options.path is None:
        dirname = os.path.dirname(options.input)
        logger.info('No path given; using %s' % dirname)
    else:
        dirname = options.path

    name = options.name

    if options.size or options.area:
        msg = 'Using deprecated options. Start using --ratio.'
        raise UserError(msg)

    learner = options.learner
    outlearner = options.outlearner

    if learner == 'No':
        ratios = (options.ratio, options.ratio)

        learn = DiffeoLearner(True, {
            'max_displ': ratios,
            'inference_method': 'sim'
        })
    else:
        logger.info('Loading diffeomorphism estimators')
        if learner[:4] == 'http':
            logger.info('from online source: %s' % learner)
            ofile = urllib.urlopen(learner)
        else:
            logger.info('from local source: %s' % learner)
            ofile = open(learner)
        learn = pickle.load(ofile)
        logger.debug('loaded')

    in_a_while = InAWhile(30)

    refine = int(options.refine)
    for _ in range(refine):
        for bagfile in args:
            i = 0
            for y0, u, y1 in read_bag(bagfile):
                logger.info('Iteration number %d' % i)
                learn.update(y0, u, y1)
                i += 1

                if in_a_while.its_time():
                    display_current_results(learn, name, dirname, i)
                    in_a_while.reset()

                if i > 15:
                    break

        if outlearner == 'No':
            pass
        else:
            logger.info('Saving learning agent to %s' % outlearner)
            pickle.dump(learn, open(outlearner, 'wb'))

        logger.info('Commands: %s' % learn.command_list)
        learn.summarize(prefix=name)

    learn.analyze(prefix=name, folder='out/diffeo-analysis/')

    learn.diffeo_dump(dirname, name)