深度学习_目标检测（二）——SSD（一）TensorFlow版本（一）使用

目录

SSD tensorflow版本

step 1/4 测试代码

1. 预训练的模型

2.执行检测

step 2/4 训练

1.voc转tfrecords数据类型

问题解决 

2.训练

问题解决 

step 3/4 验证

1.测试数据转换

2. 验证

问题解决

step 4/4 监测

参考

---

SSD tensorflow版本

step 1/4 测试代码

1. 预训练的模型

unzip ssd_300_vgg.ckpt.zip

2.执行检测

我是通过转成python脚本来运行的。官方教程如下：

jupyter notebook notebooks/ssd_notebook.ipynb

step 2/4 训练

1.voc转tfrecords数据类型

# mkdir tfrecords
DATASET_DIR=./VOC2007/
OUTPUT_DIR=./tfrecords
python tf_convert_data.py \
    --dataset_name=pascalvoc \
    --dataset_dir=./VOC2007/ \
    --output_name=voc_2007_train \
    --output_dir=./tfrecords

wiodow下

python tf_convert_data.py  --dataset_name=pascalvoc  --dataset_dir=./VOC2007/  --output_name=voc_2007_train  --output_dir=./tfrecords

问题解决 

Q: 需要更改：读取方式utf-8编码coding出错。

A: 
image_data = tf.gfile.FastGFile(filename, 'r').read()
image_data = tf.gfile.FastGFile(filename, 'rb').read()

2.训练

ssdvgg300微调

# batch_size set <16
# gpu_memory_fraction set <=0.5
```bash
DATASET_DIR=./tfrecords
TRAIN_DIR=./logs/
CHECKPOINT_PATH=./checkpoints/ssd_300_vgg.ckpt
python train_ssd_network.py \
    --train_dir=${TRAIN_DIR} \
    --dataset_dir=${DATASET_DIR} \
    --dataset_name=pascalvoc_2012 \
    --dataset_split_name=train \
    --model_name=ssd_300_vgg \
    --checkpoint_path=${CHECKPOINT_PATH} \
    --save_summaries_secs=60 \
    --save_interval_secs=600 \
    --weight_decay=0.0005 \
    --optimizer=adam \
    --learning_rate=0.001 \
    --batch_size=32
```

window下_ssdvgg300微调

```window
python train_ssd_network.py  --train_dir=./logs/  --dataset_dir=./tfrecords --dataset_name=pascalvoc_2007  --dataset_split_name=train   --model_name=ssd_300_vgg  --checkpoint_path=./checkpoints/ssd_300_vgg.ckpt  --save_summaries_secs=60  --save_interval_secs=600  --weight_decay=0.0005  --optimizer=adam  --learning_rate=0.001  --batch_size=16
```

vgg微调

DATASET_DIR=./tfrecords
TRAIN_DIR=./log/
CHECKPOINT_PATH=./checkpoints/vgg_16.ckpt
python train_ssd_network.py \
    --train_dir=${TRAIN_DIR} \
    --dataset_dir=${DATASET_DIR} \
    --dataset_name=pascalvoc_2007 \
    --dataset_split_name=train \
    --model_name=ssd_300_vgg \
    --checkpoint_path=${CHECKPOINT_PATH} \
    --checkpoint_model_scope=vgg_16 \
    --checkpoint_exclude_scopes=ssd_300_vgg/conv6,ssd_300_vgg/conv7,ssd_300_vgg/block8,ssd_300_vgg/block9,ssd_300_vgg/block10,ssd_300_vgg/block11,ssd_300_vgg/block4_box,ssd_300_vgg/block7_box,ssd_300_vgg/block8_box,ssd_300_vgg/block9_box,ssd_300_vgg/block10_box,ssd_300_vgg/block11_box \
    --trainable_scopes=ssd_300_vgg/conv6,ssd_300_vgg/conv7,ssd_300_vgg/block8,ssd_300_vgg/block9,ssd_300_vgg/block10,ssd_300_vgg/block11,ssd_300_vgg/block4_box,ssd_300_vgg/block7_box,ssd_300_vgg/block8_box,ssd_300_vgg/block9_box,ssd_300_vgg/block10_box,ssd_300_vgg/block11_box \
    --save_summaries_secs=60 \
    --save_interval_secs=600 \
    --weight_decay=0.0005 \
    --optimizer=adam \
    --learning_rate=0.001 \
    --learning_rate_decay_factor=0.94 \
    --batch_size=32

 window下_vgg微调

python train_ssd_network.py  --train_dir=./log/  --dataset_dir=./tfrecords  --dataset_name=pascalvoc_2007  --dataset_split_name=train  --model_name=ssd_300_vgg  --checkpoint_path=./checkpoints/vgg_16.ckpt  --checkpoint_model_scope=vgg_16  --checkpoint_exclude_scopes=ssd_300_vgg/conv6,ssd_300_vgg/conv7,ssd_300_vgg/block8,ssd_300_vgg/block9,ssd_300_vgg/block10,ssd_300_vgg/block11,ssd_300_vgg/block4_box,ssd_300_vgg/block7_box,ssd_300_vgg/block8_box,ssd_300_vgg/block9_box,ssd_300_vgg/block10_box,ssd_300_vgg/block11_box  --trainable_scopes=ssd_300_vgg/conv6,ssd_300_vgg/conv7,ssd_300_vgg/block8,ssd_300_vgg/block9,ssd_300_vgg/block10,ssd_300_vgg/block11,ssd_300_vgg/block4_box,ssd_300_vgg/block7_box,ssd_300_vgg/block8_box,ssd_300_vgg/block9_box,ssd_300_vgg/block10_box,ssd_300_vgg/block11_box  --save_summaries_secs=60  --save_interval_secs=600  --weight_decay=0.0005  --optimizer=adam  --learning_rate=0.001  --learning_rate_decay_factor=0.94  --batch_size=32

 

 问题解决 

Q: 无法运行训练

A:
# batch_size set <16
# gpu_memory_fraction set <=0.5

step 3/4 验证

1.测试数据转换

同理 voc转tfrecords数据类型.

# mkdir tfrecords
DATASET_DIR=./VOC2007/
OUTPUT_DIR=./test_tfrecords
python tf_convert_data.py \
    --dataset_name=pascalvoc \
    --dataset_dir=./VOC2007/ \
    --output_name=voc_2007_test \
    --output_dir=./test_tfrecords

2. 验证

EVAL_DIR=${TRAIN_DIR}/eval
python eval_ssd_network.py \
    --eval_dir=${EVAL_DIR} \
    --dataset_dir=${DATASET_DIR} \
    --dataset_name=pascalvoc_2007 \
    --dataset_split_name=test \
    --model_name=ssd_300_vgg \
    --checkpoint_path=${TRAIN_DIR} \
    --wait_for_checkpoints=True \
    --batch_size=1 \
    --max_num_batches=500

```window
python eval_ssd_network.py  --eval_dir=./logs/eval  --dataset_dir=./test_tfrecords   --dataset_name=pascalvoc_2007  --dataset_split_name=test   --model_name=ssd_300_vgg  --checkpoint_path=./logs/  --wait_for_checkpoints=True  --batch_size=1  --max_num_batches=500
```

问题解决

Q：TypeError: _variable_v2_call() got an unexpected keyword argument 'collections'

A：It should be work if you change the code in ./tf_extended/metrics.py Line 51.
"return variables.Variable" =>> "return variables.VariableV1"

step 4/4 监测

tensorboard logdir=.

参考

ssd tensorflow版本实现

https://github.com/balancap/SSD-Tensorflow

https://github.com/DJdongbudong/SSD-Tensorflow

中文教程：http://www.studyai.com/article/3e454b9e#goto-page-logo