xiaolonw / adversarial-frcnn Goto Github PK
View Code? Open in Web Editor NEWA-Fast-RCNN (CVPR 2017)
License: Other
A-Fast-RCNN (CVPR 2017)
License: Other
I want to train this net using my own dataset.
I have successfully set the data, however, the number of class need to be modified and some model layers cannot be restored accordingly.
For example, when I run the code, it raised the following error:
Cannot copy param 0 weights from layer 'cls_score_readonly'; shape mismatch. Source param shape is 21 4096 (86016); target param shape is 146 4096 (598016). To learn this layer's parameters from scratch rather than copying from a saved net, rename the layer.
Cannot copy param 0 weights from layer 'cls_score'; shape mismatch. Source param shape is 21 4096 (86016); target param shape is 146 4096 (598016). To learn this layer's parameters from scratch rather than copying from a saved net, rename the layer.
So could anybody give me some advice on how to fix it?
Thank you!
from utils.cython_bbox import bbox_overlaps
ImportError: No module named cython_bbox
I benefited from your article.
I ran into this problem while running my own data. I found the problem mainly comes from:
ASDNPretrainLabelLayer->forward()
top[0].reshape(*(conv_feat_pos.shape))
top[0].data[...] = conv_feat_pos.astype(np.float32, copy=False)
Since all the labels in rois are background (label = 0), the size of top0 is 0 when reshape top is reached.
So, how should I deal with this problem?
Thanks!
Hello, bro. Xiaolong:
I run the code for basic training of fast rcnn. However, it shown the error like this:
AssertionError: Selective search data not found at:
/home/wangxiao/Downloads/adversarial-frcnn-master/data/selective_search_data/voc_2007_trainval.mat
So, where to download this file ?
Thank you for the work.
The source of selective_search_data.tgz seems has been deleted.
The link provided in the code is
http://www.cs.berkeley.edu/~rbg/fast-rcnn-dat/selective_search_data.tgz
, which is invalid.
Can you provide a new link? Thank you.
when i implement asdn on faster-rcnn,when i run ./copy_model.h,there comes this error:
I0529 16:29:34.017501 8882 layer_factory.hpp:77] Creating layer data
I0529 16:29:34.022382 8882 net.cpp:106] Creating Layer data
I0529 16:29:34.022393 8882 net.cpp:411] data -> data
I0529 16:29:34.022400 8882 net.cpp:411] data -> im_info
I0529 16:29:34.022405 8882 net.cpp:411] data -> gt_boxes
Traceback (most recent call last):
File "python_utils/do_net_surgery.py", line 29, in
net = caffe.Net(args.out_net_def, caffe.TEST)
File "/home/keysen/lingck/py-faster-rcnn/python_utils/../lib/roi_data_layer/layer.py", line 134, in setup
top[idx].reshape(1, self._num_classes * 4)
IndexError: Index out of range
what should i do?
when i was implement ASDN on faster rcnn,there comes errors
Normalizing targets
done
RoiDataLayer: name_to_top: {'gt_boxes': 2, 'data': 0, 'im_info': 1}
AttributeError: 'module' object has no attribute 'ASDNPretrainLabelLayer'
Traceback (most recent call last):
File "tools/train_net.py", line 112, in
max_iters=args.max_iters)
File "/home/keysen/lingck/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 157, in train_net
pretrained_model=pretrained_model)
File "/home/keysen/lingck/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 43, in init
self.solver = caffe.SGDSolver(solver_prototxt)
SystemError: NULL result without error in PyObject_Call
Can you help me?
my e-mail is [email protected]
As i run the train.sh
i met this issue:
File "/home/cui/adversarial-frcnn/tools/../lib/roi_data_layer/minibatch.py", line 343, in _sample_rois
fg_inds, size=fg_rois_per_this_image, replace=False)
File "mtrand.pyx", line 1187, in mtrand.RandomState.choice (numpy/random/mtrand/mtrand.c:18864)
TypeError: 'numpy.float64' object cannot be interpreted as an index
And when i follow some answers to set the numpy to 1.11.0 it solved but appearance another issue.
And some answers tell me to upgrade the numpy .
SO, what should i do ?
Thanks very much !!!
Dear author:
I'm really interested in your work. Have you finished your whole code? Could you please share? Thank you very much.
Hello, I'm interested in your work very much, and really want to see the effect of ASTN network, so will you publish the whole code which contains ASTN & ASDN in pytorch? Thanks a lot~
In the roi_data_layer.layer.ASDNLossLayer, the bottom[0] is 'mask_pred', but in the train.prototxt, which is 'conv10_mask'. It is the slip of ASDNLossLayer or train.prototxt? Or it is right and I am confused
hi ,did the astn has been implemented ??
@xiaolonw
When I perform the training, encount the following error
Traceback (most recent call last):
File "./tools/train_net.py", line 114, in <module>
max_iters=args.max_iters)
File "/home/jmx/adversarial-frcnn/tools/../lib/fast_rcnn/train.py", line 161, in train_net
model_paths = sw.train_model(max_iters)
File "/home/jmx/adversarial-frcnn/tools/../lib/fast_rcnn/train.py", line 102, in train_model
self.solver.step(1)
File "/home/jmx/adversarial-frcnn/tools/../lib/roi_data_layer/layer.py", line 383, in forward
if min_prop > prop[j * N + i][nowlbl]:
IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`) and integer or boolean arrays are valid indices
The same error also occurred in OHEM
How do I modify it
many many thanks!!!
I want to train this adversarial frcnn network on my own dataset. But I don't know how to obtain the selective_search_data. Could anybody give me some advice? Thank you.
Hi, Xiaolong, your code interest me much. Could the idea apply on the faster rcnn code? since rpn is better to apply than selective search.
It's an amazing work, I have read your paper and it impresses me a lot.
But when I read your adversarial-frcnn/models/pascal_voc/VGG16/fast_rcnn_adv/train.prototxt , I can only find ASDN, and where is ASTN mentioned in your paper?
hi would you be able to release your code in pytorch instead of caffe please?
When I perform the training, encount the following error
Traceback (most recent call last):
File "./tools/train_net.py", line 114, in <module>
max_iters=args.max_iters)
File "/home/jmx/adversarial-frcnn/tools/../lib/fast_rcnn/train.py", line 161, in train_net
model_paths = sw.train_model(max_iters)
File "/home/jmx/adversarial-frcnn/tools/../lib/fast_rcnn/train.py", line 102, in train_model
self.solver.step(1)
File "/home/jmx/adversarial-frcnn/tools/../lib/roi_data_layer/layer.py", line 383, in forward
if min_prop > prop[j * N + i][nowlbl]:
IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`) and integer or boolean arrays are valid indices
The same error also occurred in OHEM
How do I modify it
many many thanks!!!
Thanks for the code
when i run the ./copy_model.h
it has the error , does anyone know how to solve this issue?
Thanks
I benefited from your article.
So I have tried a lot of ways to get the model and log. But it does not work.
Could tell me how to get the model and log in Chain?
Thanks!
when i was implement ASDN on faster rcnn,in the third stage,training faster_rcnn_adv,the loss of asdn was 0,could you tell me what should i change except the prototxt file?
I'm running to the last step
./experiments/scripts/fast_rcnn_adv.sh 0 VGG16 pascal_voc
This error was reported, and I didn't find text.prototxt in the folder. Do you have this file?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.