detectionteamucas / fpn_tensorflow Goto Github PK

View Code? Open in Web Editor NEW

347.0 10.0 132.0 15.34 MB

This is a tensorflow re-implementation of Feature Pyramid Networks for Object Detection.

Home Page: https://github.com/DetectionTeamUCAS/FPN_Tensorflow

License: MIT License

Python 31.22% Makefile 0.01% Jupyter Notebook 67.88% C++ 0.49% C 0.39%

fpn tensorflow object-detection

fpn_tensorflow's Introduction

Feature Pyramid Networks for Object Detection

Note

A development version based on FPN.
Support multi-gpu training!

Abstract

This is a tensorflow re-implementation of Feature Pyramid Networks for Object Detection.

This project is based on Faster-RCNN, and completed by YangXue and YangJirui.

Train on VOC 2007 trainval and test on VOC 2007 test (PS. This project also support coco training.)

Comparison

use_voc2007_metric

Models	mAP	sheep	horse	bicycle	bottle	cow	sofa	bus	dog	cat	person	train	diningtable	aeroplane	car	pottedplant	tvmonitor	chair	bird	boat	motorbike
Faster-RCNN resnet50_v1	73.09	72.11	85.63	77.74	55.82	81.19	67.34	82.44	85.66	87.34	77.49	79.13	62.65	76.54	84.01	47.90	74.13	50.09	76.81	60.34	77.47
Faster-RCNN resnet101_v1	74.63	76.35	86.18	79.87	58.73	83.4	74.75	80.03	85.4	86.55	78.24	76.07	70.89	78.52	86.26	47.80	76.34	52.14	78.06	58.90	78.04
Faster-RCNN mobilenet_v2	50.34	46.99	68.45	65.89	28.16	53.21	46.96	57.80	38.60	44.12	66.20	60.49	52.40	56.06	72.68	26.91	49.99	30.18	39.38	38.54	64.74
FPN resnet50_v1	74.26	73.27	82.23	82.99	61.27	80.59	72.73	81.37	85.26	84.76	80.33	77.43	65.31	79.18	85.78	46.47	73.10	55.99	76.11	59.80	81.19
FPN resnet101_v1	76.14	74.63	85.13	81.67	63.79	82.43	77.83	83.07	86.45	85.82	81.08	81.01	71.22	80.01	86.30	48.05	73.89	56.99	78.33	62.91	82.24
FPN resnet101_v1+	75.71	74.83	83.55	82.47	65.49	77.85	71.74	80.98	86.61	87.14	81.02	77.76	71.26	79.82	86.78	51.64	77.45	56.12	79.44	60.55	81.69
FPN resnet101_v1++	75.89	76.05	84.22	80.29	63.21	83.04	78.69	81.81	86.61	85.61	79.75	79.78	71.27	80.33	86.24	49.03	76.81	56.32	78.51	60.37	79.91

+: SHARE_NET=False
++: SHORT_SIDE_LEN=800, FAST_RCNN_MINIBATCH_SIZE=512

COCO

Model	Backbone	Train Schedule	GPU	Image/GPU	FP16	Box AP(Mask AP)
Faster (ours)	R50v1-FPN	1X	1X TITAN Xp	1	no	36.1
Faster (ours)	R50v1-FPN	1X	4X TITAN Xp	1	no	36.1
Faster (Face++ & Detectron)	R50v1-FPN	1X	8X TITAN Xp	2	no	36.4
Faster (SimpleDet)	R50v1-FPN	1X	8X 1080Ti	2	no	36.5

My Development Environment

1、python3.5 (anaconda recommend)
2、cuda9.0 (If you want to use cuda8, please set CUDA9 = False in the cfgs.py file.)
3、opencv(cv2)
4、tfplot
5、tensorflow == 1.10

Download Model

Please download resnet50_v1、resnet101_v1 pre-trained models on Imagenet, put it to $PATH_ROOT/data/pretrained_weights.

Data Format

├── VOCdevkit
│   ├── VOCdevkit_train
│       ├── Annotation
│       ├── JPEGImages
│   ├── VOCdevkit_test
│       ├── Annotation
│       ├── JPEGImages

Compile

cd $PATH_ROOT/libs/box_utils/cython_utils
python setup.py build_ext --inplace

Demo(available)

Select a configuration file in the folder ($PATH_ROOT/libs/configs/) and copy its contents into cfgs.py, then download the corresponding weights.

cd $PATH_ROOT/tools
python inference.py --data_dir='/PATH/TO/IMAGES/' 
                    --save_dir='/PATH/TO/SAVE/RESULTS/' 
                    --GPU='0'

Eval

cd $PATH_ROOT/tools
python eval.py --eval_imgs='/PATH/TO/IMAGES/'  
               --annotation_dir='/PATH/TO/TEST/ANNOTATION/'
               --GPU='0'

Train

1、If you want to train your own data, please note:

(1) Modify parameters (such as CLASS_NUM, DATASET_NAME, VERSION, etc.) in $PATH_ROOT/libs/configs/cfgs.py
(2) Add category information in $PATH_ROOT/libs/label_name_dict/lable_dict.py     
(3) Add data_name to line 76 of $PATH_ROOT/data/io/read_tfrecord.py

2、make tfrecord

cd $PATH_ROOT/data/io/  
python convert_data_to_tfrecord.py --VOC_dir='/PATH/TO/VOCdevkit/VOCdevkit_train/' 
                                   --xml_dir='Annotation'
                                   --image_dir='JPEGImages'
                                   --save_name='train' 
                                   --img_format='.jpg' 
                                   --dataset='pascal'

3、train

cd $PATH_ROOT/tools
python train.py

4、multi-gpu train

cd $PATH_ROOT/tools
python multi_gpu_train.py

Tensorboard

cd $PATH_ROOT/output/summary
tensorboard --logdir=.

Reference

1、https://github.com/endernewton/tf-faster-rcnn
2、https://github.com/zengarden/light_head_rcnn
3、https://github.com/tensorflow/models/tree/master/research/object_detection
4、https://github.com/CharlesShang/FastMaskRCNN
5、https://github.com/matterport/Mask_RCNN
6、https://github.com/msracver/Deformable-ConvNets

fpn_tensorflow's People

Contributors

Stargazers

Watchers

Forkers

kevinyang007 toubasi kmust-why styjb linden1994 tyunist masfbeca arthur-shi xuelicheng1992 tongjr1 congjianting amirunpri2018 wangshaobobetter abigeer x2ss fatsister wenbank bellaxy wenyushi jackros1022 hi-yan jbasttdi wshilin fire2323 yf longjun123456 sherrycloudy michaelbeechan qianlinjun mathpopo hxf930620 ddxx01 viibridges enyalien alexis-gruet-deel zhouyufei0001 hp13hp14 briwisdom tonygongjc dalal1983 caijiahao shenxiaojun 0fine0 zpge yusuyuan123 ybj123 guker yanghedada lewj85 clw5180 gorsun georgetzannetos bigwilky syc10-09 guohaoyuan xdboost wangxi996 allentdan deep-cognition roy699 cristal11 jrramirez3 thomastilli juniorliu95 zhaolixiang ze-bin-wu aixioma firewindmill endingback justbiubiu alexliyang sunjac henlong isxingxing hwc1824 feng-leaf x1ao-1997 wangzhe17 ysssghtz guoxiaozhong wsora xrosliang aiwenzhu qf0708443 yulilili chrisspf apulis balabala7776 wangguoning smartcai cl2227619761 murongxiaoxue zhubin1 lsun-zju luojie326 x-bruce ychuan1115 houliping chenyining123 zhaozhiyong19890102

fpn_tensorflow's Issues

pretrained weights

When I trained my model with pretrained-model, the variables are zeros like this :

model restore from pretrained mode, path is : /home/..//data/pretrained_weights/resnet_50.ckpt
resnet_v1_50/conv1/weights:0
resnet_v1_50/conv1/BatchNorm/gamma:0
resnet_v1_50/conv1/BatchNorm/beta:0
resnet_v1_50/conv1/BatchNorm/moving_mean:0
resnet_v1_50/conv1/BatchNorm/moving_variance:0
resnet_v1_50/block1/unit_1/bottleneck_v1/shortcut/weights:0
resnet_v1_50/block1/unit_1/bottleneck_v1/shortcut/BatchNorm/gamma:0
resnet_v1_50/block1/unit_1/bottleneck_v1/shortcut/BatchNorm/beta:0
resnet_v1_50/block1/unit_1/bottleneck_v1/shortcut/BatchNorm/moving_mean:0
resnet_v1_50/block1/unit_1/bottleneck_v1/shortcut/BatchNorm/moving_variance:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv1/weights:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv1/BatchNorm/gamma:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv1/BatchNorm/beta:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv1/BatchNorm/moving_mean:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv1/BatchNorm/moving_variance:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv2/weights:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv2/BatchNorm/gamma:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv2/BatchNorm/beta:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv2/BatchNorm/moving_mean:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv2/BatchNorm/moving_variance:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv3/weights:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv3/BatchNorm/gamma:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv3/BatchNorm/beta:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv3/BatchNorm/moving_mean:0
resnet_v1_50/block1/unit_1/bottleneck_v1/conv3/BatchNorm/moving_variance:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv1/weights:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv1/BatchNorm/gamma:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv1/BatchNorm/beta:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv1/BatchNorm/moving_mean:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv1/BatchNorm/moving_variance:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv2/weights:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv2/BatchNorm/gamma:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv2/BatchNorm/beta:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv2/BatchNorm/moving_mean:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv2/BatchNorm/moving_variance:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv3/weights:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv3/BatchNorm/gamma:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv3/BatchNorm/beta:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv3/BatchNorm/moving_mean:0
resnet_v1_50/block1/unit_2/bottleneck_v1/conv3/BatchNorm/moving_variance:0
resnet_v1_50/block1/unit_3/bottleneck_v1/conv1/weights:0
...
...

请问如何将输出的坐标结果映射至原始图像？

在用训练完的模型对图片进行检测时，发现输出图像的大小发生了变化，请问作者如何将输出的坐标结果映射至原始图像？

map!

你好，我把VOC2007数据集按照8:2的比例分成训练集和测试集，按照你的训练过程训练了150000次，最后得到的map为0.804，这结果与你训练测试的结果相差比较大的，我想知道，是否是我哪里出现问题了？

如何使用vgg16作为预训练模型

您好，博主，我想用vgg16代替resnet，我看networks里面有vgg.py，请问该如何修改代码

基础网络Mobilenetv2的时候训练的时候报了这个错误

Traceback (most recent call last):
File "/home/litao/Algorithm/FPN_Tensorflow-master/tools/train.py", line 188, in
train()
File "/home/litao/Algorithm/FPN_Tensorflow-master/tools/train.py", line 48, in train
gtboxes_batch=gtboxes_and_label)
File "../libs/networks/build_whole_network.py", line 390, in build_whole_detection_network
for level_name, p in zip(cfgs.LEVLES, P_list):
File "/home/litao/anaconda2/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 431, in iter
"Tensor objects are not iterable when eager execution is not "
TypeError: Tensor objects are not iterable when eager execution is not enabled. To iterate over this tensor use tf.map_fn.

train.py

Traceback (most recent call last):
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1322, in _do_call
return fn(*args)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1307, in _run_fn
options, feed_dict, fetch_list, target_list, run_metadata)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1409, in _call_tf_sessionrun
run_metadata)
tensorflow.python.framework.errors_impl.OutOfRangeError: PaddingFIFOQueue '_2_get_batch/batch/padding_fifo_queue' is closed and has insufficient elements (requested 1, current size 0)
[[Node: get_batch/batch = QueueDequeueManyV2[component_types=[DT_STRING, DT_FLOAT, DT_INT32, DT_INT32], timeout_ms=-1, _device="/job:localhost/replica:0/task:0/device:CPU:0"](get_batch/batch/padding_fifo_queue, get_batch/batch/n)]]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "/home/qzhang/Desktop/FPN_Tensorflow-master/tools/train.py", line 187, in
train()
File "/home/qzhang/Desktop/FPN_Tensorflow-master/tools/train.py", line 145, in train
_, global_stepnp = sess.run([train_op, global_step])
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 900, in run
run_metadata_ptr)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1135, in _run
feed_dict_tensor, options, run_metadata)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1316, in _do_run
run_metadata)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1335, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.OutOfRangeError: PaddingFIFOQueue '_2_get_batch/batch/padding_fifo_queue' is closed and has insufficient elements (requested 1, current size 0)
[[Node: get_batch/batch = QueueDequeueManyV2[component_types=[DT_STRING, DT_FLOAT, DT_INT32, DT_INT32], timeout_ms=-1, _device="/job:localhost/replica:0/task:0/device:CPU:0"](get_batch/batch/padding_fifo_queue, get_batch/batch/n)]]

Caused by op 'get_batch/batch', defined at:
File "/home/qzhang/Desktop/FPN_Tensorflow-master/tools/train.py", line 187, in
train()
File "/home/qzhang/Desktop/FPN_Tensorflow-master/tools/train.py", line 35, in train
is_training=True)
File "/home/qzhang/Desktop/FPN_Tensorflow-master/data/io/read_tfrecord.py", line 98, in next_batch
dynamic_pad=True)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/training/input.py", line 988, in batch
name=name)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/training/input.py", line 762, in _batch
dequeued = queue.dequeue_many(batch_size, name=name)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/data_flow_ops.py", line 483, in dequeue_many
self._queue_ref, n=n, component_types=self._dtypes, name=name)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/gen_data_flow_ops.py", line 3480, in queue_dequeue_many_v2
component_types=component_types, timeout_ms=timeout_ms, name=name)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/op_def_library.py", line 787, in _apply_op_helper
op_def=op_def)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 3392, in create_op
op_def=op_def)
File "/home/qzhang/.conda/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 1718, in init
self._traceback = self._graph._extract_stack() # pylint: disable=protected-access

OutOfRangeError (see above for traceback): PaddingFIFOQueue '_2_get_batch/batch/padding_fifo_queue' is closed and has insufficient elements (requested 1, current size 0)
[[Node: get_batch/batch = QueueDequeueManyV2[component_types=[DT_STRING, DT_FLOAT, DT_INT32, DT_INT32], timeout_ms=-1, _device="/job:localhost/replica:0/task:0/device:CPU:0"](get_batch/batch/padding_fifo_queue, get_batch/batch/n)]]

No module named 'libs.box_utils.cython_utils.cython_bbox'

你好，我在训练的时候出现这个错，然后那个cython_utils文件夹里面是没有cython_bbox文件的，请问一下这个该怎么解决啊

你好，请问config.py里面的RPN_BBOX_STD_DEV和BBOX_STD_DEV参数是如何计算的？

你好，我根据自己的数据集的特点更改了默认BASE_ANCHOR_SIZE_LIST和ANCHORS_RATIOS的参数，请问config.py里面的RPN_BBOX_STD_DEV和BBOX_STD_DEV需要相应更改吗？这两个参数是如何计算的？感谢！

The code of gtboxes is wrong. How to solve this problem? Thank you first.

image_name:b'class9_63.PNG' | rpn_loc_loss:0.00021600790205411613 | rpn_cla_loss:0.0001911183790070936 | rpn_total_loss:0.0004071262665092945 | fast_rcnn_loc_loss:0.0049437908455729485 | fast_rcnn_cla_loss:0.007616647053509951 | fast_rcnn_total_loss:0.012560438364744186 | total_loss:0.6329404711723328 | per_cost_time:0.8630495071411133s

这样的损失值正常吗，谢谢。

训练自己的数据出现的问题！

楼主你好，我用你的程序训练数据。开始我拿pascal_voc的前几十张图片测试了一下，是可以训练的。但是修改了代码之后用我自己的数据训练，格式也是pascal voc格式，出现下面这个问题

Traceback (most recent call last):
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1278, in _do_call
return fn(*args)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1263, in _run_fn
options, feed_dict, fetch_list, target_list, run_metadata)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1350, in _call_tf_sessionrun
run_metadata)
tensorflow.python.framework.errors_impl.OutOfRangeError: PaddingFIFOQueue '_1_get_batch/batch/padding_fifo_queue' is closed and has insufficient elements (requested 1, current size 0)
[[Node: get_batch/batch = QueueDequeueManyV2[component_types=[DT_STRING, DT_FLOAT, DT_INT32, DT_INT32], timeout_ms=-1, _device="/job:localhost/replica:0/task:0/device:CPU:0"](get_batch/batch/padding_fifo_queue, get_batch/batch/n)]]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "train.py", line 194, in
train()
File "train.py", line 174, in train
_, global_stepnp, summary_str = sess.run([train_op, global_step, summary_op])
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 877, in run
run_metadata_ptr)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1100, in _run
feed_dict_tensor, options, run_metadata)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1272, in _do_run
run_metadata)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/client/session.py", line 1291, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.OutOfRangeError: PaddingFIFOQueue '_1_get_batch/batch/padding_fifo_queue' is closed and has insufficient elements (requested 1, current size 0)
[[Node: get_batch/batch = QueueDequeueManyV2[component_types=[DT_STRING, DT_FLOAT, DT_INT32, DT_INT32], timeout_ms=-1, _device="/job:localhost/replica:0/task:0/device:CPU:0"](get_batch/batch/padding_fifo_queue, get_batch/batch/n)]]

Caused by op 'get_batch/batch', defined at:
File "train.py", line 194, in
train()
File "train.py", line 34, in train
is_training=True)
File "../data/io/read_tfrecord.py", line 98, in next_batch
dynamic_pad=True)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/training/input.py", line 988, in batch
name=name)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/training/input.py", line 762, in _batch
dequeued = queue.dequeue_many(batch_size, name=name)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/ops/data_flow_ops.py", line 476, in dequeue_many
self._queue_ref, n=n, component_types=self._dtypes, name=name)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/ops/gen_data_flow_ops.py", line 3480, in queue_dequeue_many_v2
component_types=component_types, timeout_ms=timeout_ms, name=name)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/framework/op_def_library.py", line 787, in _apply_op_helper
op_def=op_def)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/util/deprecation.py", line 454, in new_func
return func(*args, **kwargs)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 3155, in create_op
op_def=op_def)
File "/home/amax/panzhihao/FPN_Tensorflow/venv/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 1717, in init
self._traceback = tf_stack.extract_stack()

OutOfRangeError (see above for traceback): PaddingFIFOQueue '_1_get_batch/batch/padding_fifo_queue' is closed and has insufficient elements (requested 1, current size 0)
[[Node: get_batch/batch = QueueDequeueManyV2[component_types=[DT_STRING, DT_FLOAT, DT_INT32, DT_INT32], timeout_ms=-1, _device="/job:localhost/replica:0/task:0/device:CPU:0"](get_batch/batch/padding_fifo_queue, get_batch/batch/n)]]

我查了一下，说是数据大小不一致，但是我的数据以前在caffe上的程序跑过，并没有问题，请楼主指教一下，谢谢！

ModuleNotFoundError: No module named 'libs.box_utils.cython_utils.cython_bbox'

Traceback (most recent call last):
File "inference.py", line 17, in
from libs.networks import build_whole_network
File "../libs/networks/build_whole_network.py", line 19, in
from libs.detection_oprations.anchor_target_layer_without_boxweight import anchor_target_layer
File "../libs/detection_oprations/anchor_target_layer_without_boxweight.py", line 15, in
from libs.box_utils.cython_utils.cython_bbox import bbox_overlaps
ModuleNotFoundError: No module named 'libs.box_utils.cython_utils.cython_bbox'

您好，在训练自己的数据集过后运行eval.py进行测试，发现Precison这一指标很低，最高是0.18

在看您在cfg.py中所写的在VOC数据集上面的测试结果，Precison这一指标也不高，请问这是什么原因呢？如何操作可以提高这一值？非常感谢(｡◕◡◕｡)ﾉ

question about the version of the anaconda

您好，
请问我想用Python2.7跑这个程序能直接跑通吗，还是需要改一些文件啊？

加载模型出错

NotFoundError (see above for traceback): Key build_rpn/rpn_bbox_pred_P2/biases not found in checkpoint

cuda8.0

你好，假如说我还是想用cuda8.0的，而不想再安装一个cuda9.0来切换，该如何设置cfgs.py呢？

运行inference.py 测试时，运行的检测结果没有检测框。

您好，我们使用您提供的”voc_149999model.ckpt.data-00000-of-00001“作为权重文件，用于检测demos文件夹的测试图片。但是最终的运行结果显示，根本没有进行检测，即测试图像中不存在检测框。

或者说，您的代码是如何调用权重文件的？

no modul named tfplot

楼主，我在ubuntu系统上用pycharm创建虚拟环境打开工程，然后tensorflow-plot这个包也在配置环境里下载了，也安装成功了，但是运行demo的时候提示
File “../libs/networks/resnet.py”,line 12 ,in
ImportError: No module named tfplot
这是什么原因呢，请作者指教一下，谢谢！

使用自己的训练集训练正常，但测试出错

model restore from : /raid/cc/extend/FPN_Tensorflow/output/trained_weights/FPN_Res101_20181201/voc_120001model.ckpt
name: GeForce GTX TITAN X
major: 5 minor: 2 memoryClockRate (GHz) 1.076
pciBusID 0000:06:00.0
Total memory: 11.93GiB
Free memory: 11.81GiB
2018-12-05 17:04:54.307583: I tensorflow/core/common_runtime/gpu/gpu_device.cc:976] DMA: 0
2018-12-05 17:04:54.307590: I tensorflow/core/common_runtime/gpu/gpu_device.cc:986] 0: Y
2018-12-05 17:04:54.307597: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1045] Creating TensorFlow device (/gpu:0) -> (device: 0, name: GeForce GTX TITAN X, pci bus id: 0000:06:00.0)
restore model
2018-12-05 17:04:58.326784: I tensorflow/core/kernels/logging_ops.cc:79] SHAPE4***[0]
2018-12-05 17:04:58.326785: I tensorflow/core/kernels/logging_ops.cc:79] SHAPE3***[0]
2018-12-05 17:04:58.326818: I tensorflow/core/kernels/logging_ops.cc:79] SHAPE2***[555]
2018-12-05 17:04:58.326790: I tensorflow/core/kernels/logging_ops.cc:79] SHAPE5***[0]
/raid/cc/extend/FPN_Tensorflow/data/demos/img_6.jpg image cost 2.74012804031s:[>>>>> ]13% 1/82018-12-05 17:04:59.071122: W tensorflow/core/framework/op_kernel.cc:1192] Internal: WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
2018-12-05 17:04:59.071244: W tensorflow/core/framework/op_kernel.cc:1192] Internal: WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
[[Node: assign_levels/Where = Where_device="/job:localhost/replica:0/task:0/gpu:0"]]
2018-12-05 17:04:59.071258: W tensorflow/core/framework/op_kernel.cc:1192] Internal: WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
[[Node: assign_levels/Where = Where_device="/job:localhost/replica:0/task:0/gpu:0"]]
2018-12-05 17:04:59.071267: W tensorflow/core/framework/op_kernel.cc:1192] Internal: WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
[[Node: assign_levels/Where = Where_device="/job:localhost/replica:0/task:0/gpu:0"]]
2018-12-05 17:04:59.071294: W tensorflow/core/framework/op_kernel.cc:1192] Internal: WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
[[Node: assign_levels/Where = Where_device="/job:localhost/replica:0/task:0/gpu:0"]]
2018-12-05 17:04:59.071301: W tensorflow/core/framework/op_kernel.cc:1192] Internal: WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
[[Node: assign_levels/Where = Where_device="/job:localhost/replica:0/task:0/gpu:0"]]
2018-12-05 17:04:59.071300: W tensorflow/core/framework/op_kernel.cc:1192] Internal: WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
[[Node: assign_levels/Where = Where_device="/job:localhost/replica:0/task:0/gpu:0"]]
Traceback (most recent call last):
File "/home/cc/extend/FPN_Tensorflow/tools/test.py", line 138, in
inference_save_path=args.save_dir)
File "/home/cc/extend/FPN_Tensorflow/tools/test.py", line 105, in test
detect(det_net=faster_rcnn, inference_save_path=inference_save_path, real_test_imgname_list=test_imgname_list)
File "/home/cc/extend/FPN_Tensorflow/tools/test.py", line 60, in detect
feed_dict={img_plac: raw_img[:, :, ::-1]} # cv is BGR. But need RGB
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 895, in run
run_metadata_ptr)
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1124, in _run
feed_dict_tensor, options, run_metadata)
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1321, in _do_run
options, run_metadata)
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1340, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.InternalError: WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
[[Node: assign_levels/Where = Where_device="/job:localhost/replica:0/task:0/gpu:0"]]
[[Node: postprocess_fastrcnn/concat_2/_1287 = _Recvclient_terminated=false, recv_device="/job:localhost/replica:0/task:0/cpu:0", send_device="/job:localhost/replica:0/task:0/gpu:0", send_device_incarnation=1, tensor_name="edge_5171_postprocess_fastrcnn/concat_2", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"]]

Caused by op u'assign_levels/Where', defined at:
File "/home/cc/extend/FPN_Tensorflow/tools/test.py", line 138, in
inference_save_path=args.save_dir)
File "/home/cc/extend/FPN_Tensorflow/tools/test.py", line 105, in test
detect(det_net=faster_rcnn, inference_save_path=inference_save_path, real_test_imgname_list=test_imgname_list)
File "/home/cc/extend/FPN_Tensorflow/tools/test.py", line 35, in detect
gtboxes_batch=None)
File "/home/cc/extend/FPN_Tensorflow/libs/networks/build_whole_network.py", line 493, in build_whole_detection_network
rois_list = self.assign_levels(all_rois=rois) # rois_list: [P2_rois, P3_rois, P4_rois, P5_rois]
File "/home/cc/extend/FPN_Tensorflow/libs/networks/build_whole_network.py", line 253, in assign_levels
bbox_targets=bbox_targets)
File "/home/cc/extend/FPN_Tensorflow/libs/networks/build_whole_network.py", line 218, in get_rois
level_i_indices = tf.reshape(tf.where(tf.equal(levels, level_i)), [-1])
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/ops/array_ops.py", line 2365, in where
return gen_array_ops.where(input=condition, name=name)
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/ops/gen_array_ops.py", line 4053, in where
result = _op_def_lib.apply_op("Where", input=input, name=name)
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 767, in apply_op
op_def=op_def)
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2630, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/home/cc/miniconda3/envs/chinese-ocr/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1204, in init
self._traceback = self._graph._extract_stack() # pylint: disable=protected-access

InternalError (see above for traceback): WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
[[Node: assign_levels/Where = Where_device="/job:localhost/replica:0/task:0/gpu:0"]]
[[Node: postprocess_fastrcnn/concat_2/_1287 = _Recvclient_terminated=false, recv_device="/job:localhost/replica:0/task:0/cpu:0", send_device="/job:localhost/replica:0/task:0/gpu:0", send_device_incarnation=1, tensor_name="edge_5171_postprocess_fastrcnn/concat_2", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/cpu:0"]]

Process finished with exit code 1

我使用的是cuda8.0，还请作者赐教！

测试时导入同一个checkpoint测试结果不一样

训练完以后导入同一个checkpoint 发现每次的结果都不一样，看了下eval.py代码

sess.run(tf.global_variables_initializer())

这样导致，载入的checkpoint的数据失效，所有数据重新进行了初始化，导致模型是个随机的取值，从而每次运行获得的结果不同。

想问下您这个是不是有点小问题啊。。。。望解答谢谢~~~

feature c2

feature_dict = {'C2': end_points_C2['{}/block1/unit_2/bottleneck_v1'
C2的选择，为什么用 unit_2/bottleneck_v1？可以用unit_3吗

请问resnet.py里面的C2，C3，C4为什么跟P2，P3, P4的尺寸不一样

在Tensorboard里面看到的Layer2/C2_heat与Layer2/P2_heat的height和width不一样，请问这个C2,C3,C4具体指的是什么啊？

您好，我在12+07上训练得到的效果没有你展示效果好，这是什么原因。

恳请您的指点。

您好，有一些问题想问一下

您好，我看您的对比实验中改了SHARE_NET ，SHORT_SIDE_LEN ， FAST_RCNN_MINIBATCH_SIZE这几个参数的值，您这么改的理由是什么啊，还有就是您有发表一些相关的论文吗，想看一看

您好，在训练过程中发现真实结果的标签错位，请问是什么原因？

您好，从tensorboard中观察预测结果与真实结果的对比，发现真实结果的框出现的错位的情况，请问是什么原因呢？

RPN部分有一段代码有一个疑问

您好，我想将你们团队的代码改为检测3维目标，大概哪些文件需要修改。

训练过程出现Nan

你好，我利用您这份代码训练时出现了nan，应该不是数据的问题，因为我之前根据您另一份代码训练时成功的，tfrecord也是在那一份代码上就生成好的，但这份代码大概训练2800次以后loss就出现nan了，望解答谢谢~

Wrongs about CUDNN

I met this problem when I tried to train for my dataset.
F tensorflow/stream_executor/cuda/cuda_dnn.cc:444] could not convert BatchDescriptor {count: 0 feature_map_count: 256 spatial: 14 14 value_min: 0.000000 value_max: 0.000000 layout: BatchDepthYX} to cudnn tensor descriptor: CUDNN_STATUS_BAD_PARAM
Aborted (core dumped)

您好，我在训练过程中，在tensorboard上面观察训练结果，起初是有框的，在训练1000次左右之后，就没有框了

您好，如标题所说，在训练起初是可以看到训练框的，之后看不到训练框了，请问可能是什么原因呢？另外，layer5的热图总是看起来像是单颜色，请问这样是正常的吗？

您好，请问整体的网络结构图可以这样表示吗？

看tensorboard中的网络结构图有些看不懂，感谢您的指导~

eval.py

已经解决了，谢谢你

Download trained model by this project,but it doesn't work!

Thank you for your recommendation of the improved code，but when I replay this code,I found a problem.

@yangxue0827

大佬，你的R2CNN项目的角度预测在哪里进行的，没找到，只看到一个全连接层，多了一维输出

两个build_whole_networks有什么区别啊

关于eval的precision

为什么最后结果的precision那么小呢，我看论文中也都挺大的啊，是还要再计算什么嘛

"C:\Program Files\Anaconda3\python.exe" D:/mycode/FPN_Tensorflow-master2/tools/train.py ++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++-- D:\mycode\FPN_Tensorflow-master2 Traceback (most recent call last): File "D:/mycode/FPN_Tensorflow-master2/tools/train.py", line 16, in <module> from libs.networks import build_whole_network File "D:\mycode\FPN_Tensorflow-master2\libs\networks\build_whole_network.py", line 19, in <module> from libs.detection_oprations.anchor_target_layer_without_boxweight import anchor_target_layer File "D:\mycode\FPN_Tensorflow-master2\libs\detection_oprations\anchor_target_layer_without_boxweight.py", line 15, in <module> from libs.box_utils.cython_utils.cython_bbox import bbox_overlaps ImportError: No module named 'libs.box_utils.cython_utils.cython_bbox' Process finished with exit code 1

"C:\Program Files\Anaconda3\python.exe" D:/mycode/FPN_Tensorflow-master2/tools/train.py
++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--++--
D:\mycode\FPN_Tensorflow-master2
Traceback (most recent call last):
File "D:/mycode/FPN_Tensorflow-master2/tools/train.py", line 16, in
from libs.networks import build_whole_network
File "D:\mycode\FPN_Tensorflow-master2\libs\networks\build_whole_network.py", line 19, in
from libs.detection_oprations.anchor_target_layer_without_boxweight import anchor_target_layer
File "D:\mycode\FPN_Tensorflow-master2\libs\detection_oprations\anchor_target_layer_without_boxweight.py", line 15, in
from libs.box_utils.cython_utils.cython_bbox import bbox_overlaps
ImportError: No module named 'libs.box_utils.cython_utils.cython_bbox'

Process finished with exit code 1

在训练FPN_MobileNetv2的时候出现Tensor objects are not iterable when eager execution is not enabled

Traceback (most recent call last):
File "/home/litao/Algorithm/FPN_Tensorflow-master/tools/train.py", line 188, in
train()
File "/home/litao/Algorithm/FPN_Tensorflow-master/tools/train.py", line 48, in train
gtboxes_batch=gtboxes_and_label)
File "../libs/networks/build_whole_network.py", line 390, in build_whole_detection_network
for level_name, p in zip(cfgs.LEVLES, P_list):
File "/home/litao/anaconda2/envs/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 431, in iter
"Tensor objects are not iterable when eager execution is not "
TypeError: Tensor objects are not iterable when eager execution is not enabled. To iterate over this tensor use tf.map_fn.

你好，如果要改成Mobilenetv2,应该在你代码的基础上集中在那一部分开始改呢？

关于demo运行结果没有检测框的问题

楼主，您好，我在运行您所编辑的inference时候出现了问题
我的操作如下
1.下载resnet50_v1 到pretrained_weights文件夹下
2.在home目录下下载了voc2007的数据集（传统voc数据）
3.compile,执行了您在cython_utils文件夹下的的python setup.py build_ext --inplace命令
4.之后，我cfgs_res50_fpn.py的内容复制到cfgs.py中
5.下载权重，放置在output的trained_weights文件夹下
6.运行Demo
python inference.py --data_dir='/home/rw/FPN_Tensorflow-master/tools/demos' --save_dir='/home/rw/FPN_Tensorflow-master/tools/inference_results/1' --GPU='0'

结果如下面所示

var_in_graph: resnet_v1_50/block3/unit_3/bottleneck_v1/conv2/BatchNorm/moving_mean:0
var_in_ckpt: resnet_v1_50/block3/unit_3/bottleneck_v1/conv2/BatchNorm/moving_mean

restore from pretrained_weighs in IMAGE_NET
2019-01-22 13:24:02.512718: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2019-01-22 13:24:02.824773: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:897] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2019-01-22 13:24:02.825682: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1405] Found device 0 with properties:
name: GeForce GTX 1060 major: 6 minor: 1 memoryClockRate(GHz): 1.733
pciBusID: 0000:01:00.0
totalMemory: 5.93GiB freeMemory: 5.53GiB
2019-01-22 13:24:02.825710: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1484] Adding visible gpu devices: 0
2019-01-22 13:24:08.674217: I tensorflow/core/common_runtime/gpu/gpu_device.cc:965] Device interconnect StreamExecutor with strength 1 edge matrix:
2019-01-22 13:24:08.674294: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0
2019-01-22 13:24:08.674315: I tensorflow/core/common_runtime/gpu/gpu_device.cc:984] 0: N
2019-01-22 13:24:08.681519: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1097] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 5294 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1060, pci bus id: 0000:01:00.0, compute capability: 6.1)
restore model
/home/rw/FPN_Tensorflow-master/tools/demos/000144.jpg image cost 9.31491208076477s:[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>]100% 1/1(tf1.10)

我查看我所在的result/1这个文件夹中的图，发现只是图片大小发生了改变，然而并没有产生gif动图中的检测框，肯定是我的操作问题，请问下楼主。上面我那个地方没有做对？
这没有得到我想要的检测框，希望您能帮助下我

您好，我有几个问题，希望解答。我想问一下我训练的结果是这样子的，这会是哪里出错了呢？

数据是这样的

但训练结果却是这样的，这是为啥？

图一，这是结果，我的xml文件画出图没问题。问题一：是不是制作tfrecord的时候出错？

图二

图三，损失图

图4，

图5：问题2：这个图是什么意思？

图6：

图7：

问题3：针对这种类型的数据，如果要更好的训练，需要将图6那里设置为自己的数据吗？图7的1,和2参数应该如何改？

问题4，这个代码我训练起来很慢，大概1000张图，单1080ti，走完150000需要20小时，这正常吗？有什么提速方法吗？

非常希望您能够回答我的问题，非常感谢！

您好，我在运行eval文件的时候报了一个很奇怪的错误。

我配置了image和annotation的路径，然后运行了py文件，具体的错误如下：
InternalError (see above for traceback): WhereOp: Could not launch cub::DeviceReduce::Sum to count number of true indices. temp_storage_bytes: 1, status: invalid configuration argument
服务器配置是 cuda8+tf1.4

请问训练这个网络GUP至少要多少G

我改动了一下setup.py，使网络可以在windows上运行，但是经常运行了几步后就停止不动了，不知道是不是GPU不够的问题

does it support multiple gpus?

Skipping cancelled enqueue attempt with queue not closed

你好，我再重新跑一次这个代码的时候，在训练到9991次的时候，发现了这个错误，里面有一个警告提示是：Skipping cancelled enqueue attempt with queue not closed，紧接着后面就崩了。

@yangxue0827 请教一下大神，不知道这个问题该怎么解决的？

请问为什么运行完eval或者inference后图片的颜色变了

楼主你好，我发现运行完eval或者inference后，得到的标注后图片的颜色跟原图不一样了，请问是哪个地方造成的啊？画了锚框的图片明显比原图更亮

您好，想请问一下运行这个开源项目的顺序？

您好，我是第一次入手目标检测的项目，不是很懂里面的操作流程。按照您里面关于inference.py和eval.py的操作好像都报错了，想请问一下，是不是需要训练之后才能正确运行上述的两个py文件。然后，可以大致解释一下inference文件的作用么，是不是只是在原图里面画框？

版本问题

你好请问你使用的是tensorflow是哪一个版本，我在跑inforerence的时候出现了问题用的是1.5版的

ground truth bounding box error

When viewing with tensorboard, the groundtruth box is misplaced. Can you help me?

采用mobilenetv2报错

我在cfgs.py中设了mobilenetv2作为特征提取网络，但是报错，请问你知道是什么原因吗
Traceback (most recent call last):
File "train.py", line 186, in
train()
File "train.py", line 48, in train
gtboxes_batch=gtboxes_and_label)
File "../libs/networks/build_whole_network.py", line 383, in build_whole_detection_network
for level_name, p in zip(cfgs.LEVLES, P_list):
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/ops.py", line 431, in iter
"Tensor objects are not iterable when eager execution is not "
TypeError: Tensor objects are not iterable when eager execution is not enabled. To iterate over this tensor use tf.map_fn.