tensorflow有自己的目标检测api进行调用,仅仅在tensorflow上做了python层面上的封装,使用官方的api除了方便快捷,还有一个优点是能够直接移植到移动设备上;否则,得有些函数还需要自己在andoird上去实现,例如nms(极大值移植算法等),比较麻烦。
参考: How to train an Object Detector using Tensorflow API on Ubuntu 16.04 (GPU)
本机机器训练步骤
本地机器训练目标价检测假设以及安装上了tensorflow。
- 使用git下载model,然后配置protobuf文件,并设置环境变量
- 创建tensorflow专用数据格式进行训练。
- 配置目标检测的config文件
- 执行训练代码开始训练
制作tensorflow专有格式训练集
- 下载2012 pascal voc数据集文件并解压
wget http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar
tar -xvf VOCtrainval_11-May-2012.tar
- 生成训练以及测试数据
python object_detection/dataset_tools/create_pascal_tf_record.py \
--label_map_path=object_detection/data/pascal_label_map.pbtxt \
--data_dir=VOCdevkit --year=VOC2012 --set=train \
--output_path=pascal_train.record
python object_detection/dataset_tools/create_pascal_tf_record.py \
--label_map_path=object_detection/data/pascal_label_map.pbtxt \
--data_dir=VOCdevkit --year=VOC2012 --set=val \
--output_path=pascal_val.record
执行完后,将会生成两个tfrecord文件pascal_train.record 和 pascal_val.record。
配置目标检测的config文件
参考:Configuring the Object Detection Training Pipeline
tensorflow 目标检测api使用了protobuf配置训练以及评估过程。它能够在 object_detection/protos/pipeline.proto中找到,总的来说,配置文件被分为5个部分:
- model文件。定义了什么模型用来巡礼蓝
- train_config。它定义了训练时使用的规则(例如SGD,初始化值etc.)
- eval_config。它定义了什么值用于评估模型。
- train_input_config。它定义了训练数据集
- eval_input_config。它定义了什么数据集用于评估。
配置文件的骨架如下:
model {
(... Add model config here...)
}
train_config : {
(... Add train_config here...)
}
train_input_reader: {
(... Add train_input configuration here...)
}
eval_config: {
}
eval_input_reader: {
(... Add eval_input configuration here...)
}
- 定义输入
tensorflow接受TFRecord文件格式,用户必须声明训练以及评估文件的位置以及标签映射文件,一个例子如下:
tf_record_input_reader {
input_path: "/usr/home/username/data/train.record"
}
label_map_path: "/usr/home/username/data/label_map.pbtxt"
- 配置训练器
train_config定义了训练过程采取的规则
- 模型初始化规则
- 输入进行哪些处理
- SGD训练参数
一个简单的train_config配置如下:
batch_size: 1
optimizer {
momentum_optimizer: {
learning_rate: {
manual_step_learning_rate {
initial_learning_rate: 0.0002
schedule {
step: 0
learning_rate: .0002
}
schedule {
step: 900000
learning_rate: .00002
}
schedule {
step: 1200000
learning_rate: .000002
}
}
}
momentum_optimizer_value: 0.9
}
use_moving_average: false
}
fine_tune_checkpoint: "/usr/home/username/tmp/model.ckpt-#####"
from_detection_checkpoint: true
load_all_detection_checkpoint_vars: true
gradient_clipping_by_norm: 10.0
data_augmentation_options {
random_horizontal_flip {
}
}
训练网络模型
建议的文件训练测试架构如下所示:
+data
-label_map file
-train TFRecord file
-eval TFRecord file
+models
+ model
-pipeline config file
+train
+eval
- 训练
# From the tensorflow/models/research/ directory
PIPELINE_CONFIG_PATH={path to pipeline config file}
MODEL_DIR={path to model directory}
NUM_TRAIN_STEPS=50000
SAMPLE_1_OF_N_EVAL_EXAMPLES=1
python object_detection/model_main.py \
--pipeline_config_path=${PIPELINE_CONFIG_PATH} \
--model_dir=${MODEL_DIR} \
--num_train_steps=${NUM_TRAIN_STEPS} \
--sample_1_of_n_eval_examples=$SAMPLE_1_OF_N_EVAL_EXAMPLES \
--alsologtostderr
简单配置如下
参考:添加链接描述
# SSD with Inception v2 configuration for MSCOCO Dataset.
# Users should configure the fine_tune_checkpoint field in the train config as
# well as the label_map_path and input_path fields in the train_input_reader and
# eval_input_reader. Search for "PATH_TO_BE_CONFIGURED" to find the fields that
# should be configured.
model {
ssd {
num_classes: 1 # Set this to the number of different label classes
box_coder {
faster_rcnn_box_coder {
y_scale: 10.0
x_scale: 10.0
height_scale: 5.0
width_scale: 5.0
}
}
matcher {
argmax_matcher {
matched_threshold: 0.5
unmatched_threshold: 0.5
ignore_thresholds: false
negatives_lower_than_unmatched: true
force_match_for_each_row: true
}
}
similarity_calculator {
iou_similarity {
}
}
anchor_generator {
ssd_anchor_generator {
num_layers: 6
min_scale: 0.2
max_scale: 0.95
aspect_ratios: 1.0
aspect_ratios: 2.0
aspect_ratios: 0.5
aspect_ratios: 3.0
aspect_ratios: 0.3333
reduce_boxes_in_lowest_layer: true
}
}
image_resizer {
fixed_shape_resizer {
height: 300
width: 300
}
}
box_predictor {
convolutional_box_predictor {
min_depth: 0
max_depth: 0
num_layers_before_predictor: 0
use_dropout: false
dropout_keep_probability: 0.8
kernel_size: 3
box_code_size: 4
apply_sigmoid_to_scores: false
conv_hyperparams {
activation: RELU_6,
regularizer {
l2_regularizer {
weight: 0.00004
}
}
initializer {
truncated_normal_initializer {
stddev: 0.03
mean: 0.0
}
}
}
}
}
feature_extractor {
type: 'ssd_inception_v2' # Set to the name of your chosen pre-trained model
min_depth: 16
depth_multiplier: 1.0
conv_hyperparams {
activation: RELU_6,
regularizer {
l2_regularizer {
weight: 0.00004
}
}
initializer {
truncated_normal_initializer {
stddev: 0.03
mean: 0.0
}
}
batch_norm {
train: true,
scale: true,
center: true,
decay: 0.9997,
epsilon: 0.001,
}
}
override_base_feature_extractor_hyperparams: true
}
loss {
classification_loss {
weighted_sigmoid {
}
}
localization_loss {
weighted_smooth_l1 {
}
}
hard_example_miner {
num_hard_examples: 3000
iou_threshold: 0.99
loss_type: CLASSIFICATION
max_negatives_per_positive: 3
min_negatives_per_image: 0
}
classification_weight: 1.0
localization_weight: 1.0
}
normalize_loss_by_num_matches: true
post_processing {
batch_non_max_suppression {
score_threshold: 1e-8
iou_threshold: 0.6
max_detections_per_class: 100
max_total_detections: 100
}
score_converter: SIGMOID
}
}
}
train_config: {
batch_size: 12 # Increase/Decrease this value depending on the available memory (Higher values require more memory and vice-versa)
optimizer {
rms_prop_optimizer: {
learning_rate: {
exponential_decay_learning_rate {
initial_learning_rate: 0.004
decay_steps: 800720
decay_factor: 0.95
}
}
momentum_optimizer_value: 0.9
decay: 0.9
epsilon: 1.0
}
}
fine_tune_checkpoint: "pre-trained-model/model.ckpt" # Path to extracted files of pre-trained model
from_detection_checkpoint: true
# Note: The below line limits the training process to 200K steps, which we
# empirically found to be sufficient enough to train the pets dataset. This
# effectively bypasses the learning rate schedule (the learning rate will
# never decay). Remove the below line to train indefinitely.
num_steps: 200000
data_augmentation_options {
random_horizontal_flip {
}
}
data_augmentation_options {
ssd_random_crop {
}
}
}
train_input_reader: {
tf_record_input_reader {
input_path: "annotations/train.record" # Path to training TFRecord file
}
label_map_path: "annotations/label_map.pbtxt" # Path to label map file
}
eval_config: {
num_examples: 8000
# Note: The below line limits the evaluation process to 10 evaluations.
# Remove the below line to evaluate indefinitely.
max_evals: 10
}
eval_input_reader: {
tf_record_input_reader {
input_path: "annotations/test.record" # Path to testing TFRecord
}
label_map_path: "annotations/label_map.pbtxt" # Path to label map file
shuffle: false
num_readers: 1
}