smallcorgi / Faster-RCNN_TF

Faster-RCNN in Tensorflow
MIT License
2.34k stars 1.12k forks source link

run demo.py (core dumped) #206

Open gentlebreeze1 opened 7 years ago

gentlebreeze1 commented 7 years ago

when i run ./experiments/scripts/faster_rcnn_end2end.sh gpu 0 VGG16 pascal_voc the problem occurs as follow:faster_rcnn_end2end.sh: line 57: 24466 Segmentation fault (core dumped) python ./tools/train_net.py --device ${DEV} --device_id ${DEV_ID} --weights data/pretrain_model/VGG_imagenet.npy --imdb ${TRAIN_IMDB} --iters ${ITERS} --cfg experiments/cfgs/faster_rcnn_end2end.yml --network VGGnet_train ${EXTRA_ARGS}

thank you

gentlebreeze1 commented 7 years ago

someone help

Akansha06 commented 7 years ago

Hi @gentlebreeze1 that is because you are using a larger network on small cores of GPU. Please check your configuration once.

BrownOfSummer commented 7 years ago

Faild on CPU only when run the demo, test on tensor flow 1.0, 1.2, 1.3: Tensor("Placeholder:0", shape=(?, ?, ?, 3), dtype=float32) Tensor("conv5_3/conv5_3:0", shape=(?, ?, ?, 512), dtype=float32) Tensor("rpn_conv/3x3/rpn_conv/3x3:0", shape=(?, ?, ?, 512), dtype=float32) Tensor("rpn_cls_score/rpn_cls_score:0", shape=(?, ?, ?, 18), dtype=float32) Tensor("rpn_cls_prob:0", shape=(?, ?, ?, ?), dtype=float32) Tensor("rpn_cls_prob_reshape:0", shape=(?, ?, ?, 18), dtype=float32) Tensor("rpn_bbox_pred/rpn_bbox_pred:0", shape=(?, ?, ?, 36), dtype=float32) Tensor("Placeholder_1:0", shape=(?, 3), dtype=float32) Tensor("conv5_3/conv5_3:0", shape=(?, ?, ?, 512), dtype=float32) Tensor("rois:0", shape=(?, 5), dtype=float32) [<tf.Tensor 'conv5_3/conv5_3:0' shape=(?, ?, ?, 512) dtype=float32>, <tf.Tensor 'rois:0' shape=(?, 5) dtype=float32>] Tensor("fc7/fc7:0", shape=(?, 4096), dtype=float32)

Loaded network /home/VGGnet_fast_rcnn_iter_70000.ckpt Segmentation fault (core dumped)

Akansha06 commented 7 years ago

Hi @BrownOfSummer,

Please share your hardware configurations once

gentlebreeze1 commented 7 years ago

i get result as follow: AP for "wave one hand" = -1.0000 AP for punch = -1.0000 AP for "wave two hand" = -1.0000 AP for "sit down" = -1.0000 AP for "pick up" = -1.0000 AP for clap = -1.0000 AP for walk = -1.0000 AP for nod = -1.0000 AP for jump = -1.0000 Mean AP = -1.0000


Results:
-1.000
-1.000
-1.000
-1.000
-1.000
-1.000
-1.000
-1.000
-1.000
-1.000
what the meaning of -1? @Akansha06 
gentlebreeze1 commented 7 years ago

if i want to use Alxnet training ,what should i change? @Akansha06

Nofcity commented 7 years ago

@BrownOfSummer,hi, I met the same error,Segmentation fault (core dumped),have you solved it?Is it not enough in CPU memory?thanks!!!

xulzee commented 6 years ago

@Nofcity @Nofcity hi, I met the same error ,Segmentation fault(core dumped) Is it not enough in CPU memory?thanks!!!

ChampagneLee commented 6 years ago

Hi @BrownOfSummer, I met this error too. Have you solved it? Thanks!

hzylmf commented 6 years ago

I got the same error when running demo.py or train_net.py, i can make sure that the error occur in: rpn_loss_cls_value, rpn_loss_box_value,loss_cls_value, loss_boxvalue, = sess.run([rpn_cross_entropy, rpn_loss_box, cross_entropy, loss_box, train_op], feed_dict=feed_dict, options=run_options, run_metadata=run_metadata) so, how to solve it?

amlandas78 commented 6 years ago

tee: experiments/logs/faster_rcnn_end2endVGG16.txt.2018-04-24_23-11-57: No such file or directory

Can someone give me solution regarding this error?

hzylmf commented 6 years ago

@amlandas78 hi, I have no idea about how to solve the segmentation fault, but I think you can try this implement:tf-faster-rcnn, it works well.

amlandas78 commented 6 years ago

@hzylmf @Akansha06 @gentlebreeze1 : Is this due to the reason that I am using a gpu with a very low configuration..Its is just a geforce 920M laptop gpu with 4gb ram with cuda compute capability of 3.5?

hzylmf commented 6 years ago

@amlandas78 Maybe~ your gpu is too low...but I got segmentation fault too when I used gtx1080ti. So you'd better try another implement than debug it if you can not find the deep reason.

gentlebreeze1 commented 6 years ago

File "/media/sys51/0002488400013A3B/students/yongzhang/Faster-RCNN_TF/tools/../lib/datasets/pascal_voc.py", line 212, in _load_pascal_annotation cls = self._class_to_ind[obj.find('name').text.lower().strip()] KeyError: u'\u963f\u82ac\u773c\u8776' my name is 'aa'and 'ab' what's wrong? someone help? @Akansha06 @hzylmf

hzylmf commented 6 years ago

@gentlebreeze1 It seems that there are chinese characters in your samples, or you are using python2 instead of python3?

gentlebreeze1 commented 6 years ago

my python version is python2. Python 2.7.12 (default, Dec 4 2017, 14:50:18) [GCC 5.4.0 20160609] on linux2 Type "help", "copyright", "credits" or "license" for more information.

@hzylmf

gentlebreeze1 commented 6 years ago
hudie"IMG_000540.jpg"C:\Users\Ji\DesktopUnknown6000400030abUnspecified002632179733812483 @hzylmf
gentlebreeze1 commented 6 years ago

The result does not converge.what method can solve the problem ?Reduce learning rate? @Akansha06 @hzylmf

hxf930620 commented 6 years ago

@gentlebreeze1 do you know why AP=-1?