ResNet 修改

https://github.com/tornadomeet/ResNet

apache 开源项目

修改如下：

训练模块

import argparse,logging,os
import mxnet as mx
from symbol_resnet import resnet

logger = logging.getLogger()
logger.setLevel(logging.INFO)

def multi_factor_scheduler(begin_epoch, epoch_size, step=[60, 75, 90], factor=0.1):
    step_ = [epoch_size * (x-begin_epoch) for x in step if x-begin_epoch > 0]
    return mx.lr_scheduler.MultiFactorScheduler(step=step_, factor=factor) if len(step_) else None

def main():
    if args.data_type == "cifar10":
        args.aug_level = 1
        args.num_classes = 10
        # depth should be one of 110, 164, 1001,...,which is should fit (args.depth-2)%9 == 0
        if((args.depth-2)%9 == 0 and args.depth >= 164):
            per_unit = [(args.depth-2)/9]
            filter_list = [16, 64, 128, 256]
            bottle_neck = True
        elif((args.depth-2)%6 == 0 and args.depth < 164):
            per_unit = [(args.depth-2)/6]
            filter_list = [16, 16, 32, 64]
            bottle_neck = False
        else:
            raise ValueError("no experiments done on detph {}, you can do it youself".format(args.depth))
        units = per_unit*3
        symbol = resnet(units=units, num_stage=3, filter_list=filter_list, num_class=args.num_classes,
                        data_type="cifar10", bottle_neck = bottle_neck, bn_mom=args.bn_mom, workspace=args.workspace,
                        memonger=args.memonger)
    elif args.data_type == "imagenet":
        args.num_classes = 3
        if args.depth == 18:
            units = [2, 2, 2, 2]
        elif args.depth == 34:
            units = [3, 4, 6, 3]
        elif args.depth == 50:
            units = [3, 4, 6, 3]
        elif args.depth == 101:
            units = [3, 4, 23, 3]
        elif args.depth == 152:
            units = [3, 8, 36, 3]
        elif args.depth == 200:
            units = [3, 24, 36, 3]
        elif args.depth == 269:
            units = [3, 30, 48, 8]
        else:
            raise ValueError("no experiments done on detph {}, you can do it youself".format(args.depth))
        symbol = resnet(units=units, num_stage=4, filter_list=[64, 256, 512, 1024, 2048] if args.depth >=50
                        else [64, 64, 128, 256, 512], num_class=args.num_classes, data_type="imagenet", bottle_neck = True
                        if args.depth >= 50 else False, bn_mom=args.bn_mom, workspace=args.workspace,
                        memonger=args.memonger)
    else:
         raise ValueError("do not support {} yet".format(args.data_type))
    kv = mx.kvstore.create(args.kv_store)
    devs = mx.cpu() if args.gpus is None else [mx.gpu(int(i)) for i in args.gpus.split(',')]
    epoch_size = max(int(args.num_examples / args.batch_size / kv.num_workers), 1)
    begin_epoch = args.model_load_epoch if args.model_load_epoch else 0
    if not os.path.exists("./model"):
        os.mkdir("./model")
    model_prefix = "model/resnet-{}-{}-{}".format(args.data_type, args.depth, kv.rank)
    checkpoint = mx.callback.do_checkpoint(model_prefix)
    arg_params = None
    aux_params = None
    if args.retrain:
        _, arg_params, aux_params = mx.model.load_checkpoint(model_prefix, args.model_load_epoch)
    if args.memonger:
        import memonger
        symbol = memonger.search_plan(symbol, data=(args.batch_size, 3, 32, 32) if args.data_type=="cifar10"
                                                    else (args.batch_size, 3, 128, 128))
    train = mx.io.ImageRecordIter(
        path_imgrec         = os.path.join(args.data_dir, "cifar10_train.rec") if args.data_type == 'cifar10' else
                              os.path.join(args.data_dir, "train_256_q90.rec") if args.aug_level == 1
                              else os.path.join(args.data_dir, "train_480_q90.rec"),
        label_width         = 1,
        data_name           = 'data',
        label_name          = 'softmax_label',
        data_shape          = (3, 32, 32) if args.data_type=="cifar10" else (3, 128, 128),
        batch_size          = args.batch_size,
        pad                 = 4 if args.data_type == "cifar10" else 0,
        fill_value          = 127,  # only used when pad is valid
        rand_crop           = True,
        max_random_scale    = 1.0,  # 480 with imagnet, 32 with cifar10
        min_random_scale    = 1.0 if args.data_type == "cifar10" else 1.0 if args.aug_level == 1 else 0.533,  # 256.0/480.0
        max_aspect_ratio    = 0 if args.data_type == "cifar10" else 0 if args.aug_level == 1 else 0.25,
        random_h            = 0 if args.data_type == "cifar10" else 0 if args.aug_level == 1 else 36,  # 0.4*90
        random_s            = 0 if args.data_type == "cifar10" else 0 if args.aug_level == 1 else 50,  # 0.4*127
        random_l            = 0 if args.data_type == "cifar10" else 0 if args.aug_level == 1 else 50,  # 0.4*127
        max_rotate_angle    = 0 if args.aug_level <= 2 else 10,
        max_shear_ratio     = 0 if args.aug_level <= 2 else 0.1,
        rand_mirror         = True,
        shuffle             = True,
        num_parts           = kv.num_workers,
        part_index          = kv.rank)
    val = mx.io.ImageRecordIter(
        path_imgrec         = os.path.join(args.data_dir, "cifar10_val.rec") if args.data_type == 'cifar10' else
                              os.path.join(args.data_dir, "val_256_q90.rec"),
        label_width         = 1,
        data_name           = 'data',
        label_name          = 'softmax_label',
        batch_size          = args.batch_size,
        data_shape          = (3, 32, 32) if args.data_type=="cifar10" else (3, 128, 128),
        rand_crop           = False,
        rand_mirror         = False,
        num_parts           = kv.num_workers,
        part_index          = kv.rank)
    model = mx.model.FeedForward(
        ctx                 = devs,
        symbol              = symbol,
        arg_params          = arg_params,
        aux_params          = aux_params,
        num_epoch           = 200 if args.data_type == "cifar10" else 120,
        begin_epoch         = begin_epoch,
        learning_rate       = args.lr,
        momentum            = args.mom,
        wd                  = args.wd,
        optimizer           = 'nag',
        # optimizer          = 'sgd',
        initializer         = mx.init.Xavier(rnd_type='gaussian', factor_type="in", magnitude=2),
        lr_scheduler        = multi_factor_scheduler(begin_epoch, epoch_size, step=[120, 160], factor=0.1)
                             if args.data_type=='cifar10' else
                             multi_factor_scheduler(begin_epoch, epoch_size, step=[30, 60, 90], factor=0.1),
        )
    model.fit(
        X                  = train,
        eval_data          = val,
        eval_metric        = ['acc', 'ce'] if args.data_type=='cifar10' else
                             ['acc','ce', mx.metric.create('top_k_accuracy', top_k = 5)],
        kvstore            = kv,
        batch_end_callback = mx.callback.Speedometer(args.batch_size, args.frequent),
        epoch_end_callback = checkpoint)
    # logging.info("top-1 and top-5 acc is {}".format(model.score(X = val,
    #               eval_metric = ['acc', mx.metric.create('top_k_accuracy', top_k = 5)])))

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="command for training resnet-v2")
    parser.add_argument('--gpus', type=str, default='', help='the gpus will be used, e.g "0,1,2,3"')
    parser.add_argument('--data-dir', type=str, default='./data/imagenet/', help='the input data directory')
    parser.add_argument('--data-type', type=str, default='imagenet', help='the dataset type')
    parser.add_argument('--list-dir', type=str, default='./',
                        help='the directory which contain the training list file')
    parser.add_argument('--lr', type=float, default=0.1, help='initialization learning reate')
    parser.add_argument('--mom', type=float, default=0.9, help='momentum for sgd')
    parser.add_argument('--bn-mom', type=float, default=0.9, help='momentum for batch normlization')
    parser.add_argument('--wd', type=float, default=0.0001, help='weight decay for sgd')
    parser.add_argument('--batch-size', type=int, default=256, help='the batch size')
    parser.add_argument('--workspace', type=int, default=512, help='memory space size(MB) used in convolution, if xpu '
                        ' memory is oom, then you can try smaller vale, such as --workspace 256')
    parser.add_argument('--depth', type=int, default=50, help='the depth of resnet')
    parser.add_argument('--num-classes', type=int, default=1000, help='the class number of your task')
    parser.add_argument('--aug-level', type=int, default=2, choices=[1, 2, 3],
                        help='level 1: use only random crop and random mirror\n'
                             'level 2: add scale/aspect/hsv augmentation based on level 1\n'
                             'level 3: add rotation/shear augmentation based on level 2')
    parser.add_argument('--num-examples', type=int, default=1281167, help='the number of training examples')
    parser.add_argument('--kv-store', type=str, default='device', help='the kvstore type')
    parser.add_argument('--model-load-epoch', type=int, default=0,
                        help='load the model on an epoch using the model-load-prefix')
    parser.add_argument('--frequent', type=int, default=50, help='frequency of logging')
    parser.add_argument('--memonger', action='store_true', default=False,
                        help='true means using memonger to save momory, https://github.com/dmlc/mxnet-memonger')
    parser.add_argument('--retrain', action='store_true', default=False, help='true means continue training')
    args = parser.parse_args()
    logging.info(args)
    main()

为减小网络大小，将图片全部缩放为128*128大小，平时使用ResNet-50的网络，将num_classes 改为需要的分类数目。

train acc可以在99.9%水平，val acc 稳定在80%左右

预测模块

import numpy as np
import cv2
import mxnet as mx
import argparse

def ch_dev(arg_params, aux_params, ctx):
    new_args = dict()
    new_auxs = dict()
    for k, v in arg_params.items():
        new_args[k] = v.as_in_context(ctx)
    for k, v in aux_params.items():
        new_auxs[k] = v.as_in_context(ctx)
    return new_args, new_auxs

def predict(img):
    # compute the predict probabilities
    mod.forward(Batch([img]))
    prob = mod.get_outputs()[0].asnumpy()
    # print the top-5
    prob = np.squeeze(prob)
    a = np.argsort(prob)[::-1]
    for i in a[0:3]:
        print('probability=%f, class=%s' %(prob[i], labels[i]))

def main():
    synset = [l.strip() for l in open(args.synset).readlines()]
    # 添加预测
    ctx = mx.gpu(args.gpu)
    sym, arg_params, aux_params = mx.model.load_checkpoint(args.prefix, args.epoch)
    mod = mx.mod.Module(symbol=sym, context=ctx, label_names=None)
    mod.bind(for_training=False, data_shapes=[('data', (1,3,128,128))],label_shapes=mod._label_shapes)
    mod.set_params(arg_params, aux_params, allow_missing=True)

    from collections import namedtuple
    Batch = namedtuple('Batch', ['data'])

    if args.lst:
        file = open('instances_test.lst')
        for line in file:
            src = ""
            for i in range(len(line)-1,0,-1):
                if line[i] == '\t':
                    break
                src += line[i]
            src = src[::-1]
            src = "/mnt/hdfs-data-4/data/jian.yin/ped_thumbnail/instances_test/" + src
            print(src[0:-1])

            # convert into format (batch, RGB, width, height)
            img = mx.image.imdecode(open(src[0:-1],'rb').read())
            img = mx.image.imresize(img, 128, 128) # resize
            img = img.transpose((2, 0, 1)) # Channel first
            img = img.expand_dims(axis=0) # batchify
            img = img.astype('float32') # for gpu context

            mod.forward(Batch([img]))
            prob = mod.get_outputs()[0].asnumpy()
            # print the top-3
            prob = np.squeeze(prob)
            a = np.argsort(prob)[::-1]
            for i in a[0:3]:
                print('probability=%f, class=%s' %(prob[i], synset[i]))

#             img = cv2.cvtColor(cv2.imread(src[0:-1]), cv2.COLOR_BGR2RGB)
#             img = cv2.resize(img, (128, 128))  # resize to 224*224 to fit model
#             img = np.swapaxes(img, 0, 2)
#             img = np.swapaxes(img, 1, 2)  # change to (c, h,w) order
#             img = img[np.newaxis, :]  # extend to (n, c, h, w)

#             ctx = mx.gpu(args.gpu)
#             sym, arg_params, aux_params = mx.model.load_checkpoint(args.prefix, args.epoch)
#             arg_params, aux_params = ch_dev(arg_params, aux_params, ctx)
#             arg_params["data"] = mx.nd.array(img, ctx)
#             arg_params["softmax_label"] = mx.nd.empty((1,), ctx)
#             exe = sym.bind(ctx, arg_params ,args_grad=None, grad_req="null", aux_states=aux_params)
#             exe.forward(is_train=False)

#             prob = np.squeeze(exe.outputs[0].asnumpy())
#             pred = np.argsort(prob)[::-1]
#             print("Top1 result is: ", synset[pred[0]])
#             # print("Top5 result is: ", [synset[pred[i]] for i in range(5)])
        file.close()

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="use pre-trainned resnet model to classify one image")
    parser.add_argument('--img', type=str, default='test.jpg', help='input image for classification')
    # add --lst
    parser.add_argument('--lst',type=str,default='test.lst',help="input image's lst for classification")
    parser.add_argument('--gpu', type=int, default=0, help='the gpu id used for predict')
    parser.add_argument('--synset', type=str, default='synset.txt', help='file mapping class id to class name')
    parser.add_argument('--prefix', type=str, default='resnet-50', help='the prefix of the pre-trained model')
    parser.add_argument('--epoch', type=int, default=0, help='the epoch of the pre-trained model')
    args = parser.parse_args()
    main()

添加了--lst可选参数，可以批处理序列化文件预测。

原文预测模块效率较低，改用mxnet标准的predict写法：https://mxnet.incubator.apache.org/tutorials/python/predict_image.html

添加一个脚本，防止忘记一些参数的写法：

#!/usr/bin/
python -u predict.py --lst instances_test.lst --prefix resnet-50 --synset ped_thumbnail.txt --gpu 0

记得运行的时候添加管道命令 >

/mnt/1/385_328_428_402_6.jpg
probability=0.994927, class=1 Cyclist
probability=0.003335, class=2 Others
probability=0.001739, class=0 Pedestrian
/mnt2/439_359_481_428_0.jpg
probability=0.994793, class=2 Others
probability=0.002817, class=0 Pedestrian
probability=0.002390, class=1 Cyclist
/mnt/2/619_337_658_401_16.jpg
probability=0.992218, class=2 Others
probability=0.007275, class=1 Cyclist
probability=0.000507, class=0 Pedestrian
/mnt1/511_288_561_385_1.jpg
probability=0.997837, class=1 Cyclist
probability=0.001525, class=0 Pedestrian
probability=0.000638, class=2 Others

分析预测结果

可以先把各种分类的路径记录下来。

import itertools
import numpy as np
import matplotlib.pyplot as plt

from sklearn import svm, datasets
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix

file = open('myPredict.txt')

cnt = 0

true = []
pred = []

for line in file:
    if cnt%4 == 0:
        pos =-1
        for i in range(len(line)-1,-1,-1):
            if line[i]=='/':
                pos = i - 1
                break
        true.append(int(line[pos]))
    if cnt%4 == 1:
        pos = -1
        for i in range(len(line)-1,-1,-1):
            if line[i] == ' ':
                pos = i - 1
                break
        pred.append(int(line[pos]))
    cnt+=1

print(true)
print(pred)
print(confusion_matrix(true,pred))

file.close()

zero_zero = []
zero_one = []
zero_two = []
one_zero = []
one_one = []
one_two = []
two_zero = []
two_one = []
two_two = []

cnt = 0
pos = 0
file = open('myPredict.txt')

for line in file:
    if cnt%4==0:
        if true[pos] == 0 and pred[pos] == 0:
            zero_zero.append(line)
        if true[pos] == 0 and pred[pos] == 1:
            zero_one.append(line)
        if true[pos] == 0 and pred[pos] == 2:
            zero_two.append(line)
        if true[pos] == 1 and pred[pos] == 0:
            one_zero.append(line)
        if true[pos] == 1 and pred[pos] == 1:
            one_one.append(line)
        if true[pos] == 1 and pred[pos] == 2:
            one_two.append(line)
        if true[pos] == 2 and pred[pos] == 0:
            two_zero.append(line)
        if true[pos] == 2 and pred[pos] == 1:
            two_one.append(line)
        if true[pos] == 2 and pred[pos] == 2:
            two_two.append(line)
        pos+=1
    cnt+=1
file.close()

print(len(zero_one)+len(zero_two)+len(one_zero)+len(one_two)+len(two_zero)+len(two_one))

# 0 - 0
write_zero_zero = open('zero_zero.txt','w')
for i in range(len(zero_zero)):
    write_zero_zero.write(zero_zero[i])
write_zero_zero.close()

# 0 - 1
write_zero_one = open('zero_one.txt','w')
for i in range(len(zero_one)):
    write_zero_one.write(zero_one[i])
write_zero_one.close()

# 0 - 2
write_zero_two = open('zero_two.txt','w')
for i in range(len(zero_two)):
    write_zero_two.write(zero_two[i])
write_zero_two.close()

# 1 - 0
write_one_zero = open('one_zero.txt','w')
for i in range(len(one_zero)):
    write_one_zero.write(one_zero[i])
write_one_zero.close()

# 1 - 1
write_one_one = open('one_one.txt','w')
for i in range(len(one_one)):
    write_one_one.write(one_one[i])
write_one_one.close()

# 1 - 2
write_one_two = open('one_two.txt','w')
for i in range(len(one_two)):
    write_one_two.write(one_two[i])
write_one_two.close()

# 2 - 0
write_two_zero = open('two_zero.txt','w')
for i in range(len(two_zero)):
    write_two_zero.write(two_zero[i])
write_two_zero.close()

# 2 - 1
write_two_one = open('two_one.txt','w')
for i in range(len(two_one)):
    write_two_one.write(two_one[i])
write_two_one.close()

# 2 - 2
write_two_two = open('two_two.txt','w')
for i in range(len(two_two)):
    write_two_two.write(two_two[i])
write_two_two.close()

混淆矩阵如下：