py-faster-rcnn on new dataset

[leefionglee]

Hi all: I would like to train py-faster-rcnn on my own dataset, but what is exactly the data format? i.e. the images, annotations, train, val? Can anyone post an example of annotation file here? As I found one post here: https://github.com/zeyuanxy/fast-rcnn/tree/master/help/train, in which the annotation is text, while I also found PASCAL is XML. What exactly it is? Any reference blogs or tutorials will be highly appreciated. Thanks

[athus1990]

pascal has an xml file...look at the original VOCdevkit code(readme and helper function)...Xml basically can be converted to a struct in matlab and u can see that it has information about the class name,bounding boxes etc..

[yileo19920925]

hi~ @leefionglee I trained fastrcnn on other dataset lastweek~ weather xml or txt in annotation does not matter. you can see in path /fast-rcnn/lib/datasets/inria.py there is a func named _load_inria_annotation to deal with annotation file :) while i used xml get annotation

[kshalini]

@yileo19920925 can you please share some insights on your training steps. is it as simple as getting the data in the format of VOC2007 trainval and test, and placing them into the right folder paths and just calling alt_opt.sh?

do we need to write any python code or modify any other files like FastrCNN training (factory.py etc?) thanks in advance

[yileo19920925]

@kshalini hi I trained my datasets using fast(not er) r cnn..... BUT i think whether fast or faster rcnn require imdb to make train process In order to deal with imdb we need to write or modify pascal_voc.py The function of factory.py is to add this kind of datasets like pascal_voc

when i trained my data .It helped me a lot (i hope it will be positive to u) https://github.com/coldmanck/fast-rcnn/blob/master/README.md

about alt_opt. i did not used it when trained In fast rcnn u can use ./tools/train_net.py In faster rcnn ./tool/train_faster_rcnn_alt_opt.py may be more straightforward

[leejiajun]

@leefionglee @athus1990 @kshalini

i trained with ImageNet, but met an error in /lib/rpn/anchor_target_layer.py.

#error
Process Process-1:
Traceback (most recent call last):
  File "/usr/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap
    self.run()
  File "/usr/lib/python2.7/multiprocessing/process.py", line 114, in run
    self._target(*self._args, **self._kwargs)
  File "./tools/train_faster_rcnn_alt_opt.py", line 132, in train_rpn
    max_iters=max_iters)
  File "/home/lijiajun/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 136, in train_net
    model_paths = sw.train_model(max_iters)
  File "/home/lijiajun/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 104, in train_model
    self.solver.step(1)
  File "/home/lijiajun/py-faster-rcnn/tools/../lib/rpn/anchor_target_layer.py", line 137, in forward
    gt_argmax_overlaps = overlaps.argmax(axis=0)
ValueError: attempt to get argmax of an empty sequence

when run at "only keep anchors inside the image", len(ids_inside) is equal to 0.

    # only keep anchors inside the image
    inds_inside = np.where(
        (all_anchors[:, 0] >= -self._allowed_border) &
        (all_anchors[:, 1] >= -self._allowed_border) &
        (all_anchors[:, 2] < im_info[1] + self._allowed_border) &  # width
        (all_anchors[:, 3] < im_info[0] + self._allowed_border)    # height
    )[0]

how to solver?

[andrewliao11]

@leefionglee @kshalini Hi! I've train faster rcnn on imagenet (200 categories), hope this can help you! https://github.com/andrewliao11/py-faster-rcnn/blob/master/README.md

[banxiaduhuo]

@leejiajun Hi, leejiajun, Did you solve this problem？I have the same problem.

[LiberiFatali]

@leejiajun @banxiaduhuo Hi, did you solve the problem? I got the same issue when training ZF model on custom dataset

[leejiajun]

@banxiaduhuo @LiberiFatali

Because, the ratio of images width and height is too small or large. You could remove that images and solve this problem.

[LiberiFatali]

Right, some images have big ratio of width and height cause this.Thanks!

[ck196]

@leejiajun How do you train imagenet dataset? Could you write some instructions for this step. Thank you.

[leejiajun]

@monkeykju all work focus on coding lib/datasets/pascal_voc.py and lib/datasets/factory.py you could read this, https://github.com/andrewliao11/py-faster-rcnn/blob/master/README.md

[daf11865]

@leefionglee can you tell me why the training image ratio would cause such an error? is that because RPN produce no roi which can be considered gt under that ratio? if yes, why too big or too small image would effect this thank you.

[karenyun]

@leefionglee I do as the inria.py(two classes), but when i train my dataset, the machine always crash, i forget to prepare the negative images, this will cause the issue or there may other reason? such as: How to make the negative images and annotation? thanks.

[LiberiFatali]

I remember that this py-faster-rccn doesn't used selective search

[aragon111]

@MarkoArsenovic I wanna train the net on my own dataset. I don't understand how to modify the python files before training. I can't find any repo where those steps are explained.

[deboc]

Hi, Let's have a look there

[aragon111]

Thanks @deboc . I used coldmanck instructions for training the fast rcnn. The problem is that there are few steps which are differents (in the construct of the IMDB file and editing the shell script)

[deboc]

Ok tools/train_net.py is in fact an old script for fast-rcnn. You can directly launch the training without any script : (example for alt_opt training) $cd <py-faster-rcnn folder> $./tools/train_faster_rcnn_alt_opt.py --gpu 0 --net_name <model name> --weights <pretrained .caffemodel> --imdb <dataset name>_train

or just use tools/train_faster_rcnn_alt_opt.py

[deboc]

Ok we definitely miss a real tutorial dedicated to py-faster-rcnn with a simple dataset. Chances are your inria.py need 2 more methods (rpn_roidb & _load_rpn_roidb). I'm sorry not to remember how exactly I built mine, but if you want to finish the job have a look here

[aragon111]

Hello @deboc , I tried running the training as you explained but after I typed

./train_faster_rcnn_alt_opt.py --gpu 0 --solver /models/pascal_voc/ZF/faster_rcnn_end2end/solver.prototxt --weights /data/faster_rcnn_models/ZF_faster_rcnn_final.caffemodel --imdb amphora_train

I got this error: train_faster_rcnn_alt_opt.py: error: unrecognized arguments: --solver /models/pascal_voc/ZF/faster_rcnn_end2end/solver.prototxt

I don't understand what is wrong with the solver I chose.

[deboc]

You are using the alt_opt script with a end2end solver, so it won't work.

For training alt_opt the argument is not --solver <solver.prototxt> but --net_name <model_name> which specify the model folder. Then train_faster_rcnn_alt_opt.py automatically looks for solvers in models/<model_name>/faster_rcnn_alt_opt/ where you should put solvers for alt_opt training.

[aragon111]

Thank you @deboc . I understood now and I edited the faster_rcnn_test file as I have done in fast-rcnn. I got now an error about the stage1_rpn_train file (which is in the right directory).

I0421 19:04:00.501624 20093 solver.cpp:61] Creating training net from train_net file: models/pascal_voc/ZF/faster_rcnn_alt_opt/stage1_rpn_train.pt F0421 19:04:00.501646 20093 io.cpp:34] Check failed: fd != -1 (-1 vs. -1) File not found: models/pascal_voc/ZF/faster_rcnn_alt_opt/stage1_rpn_train.pt

[aragon111]

@MarkoArsenovic did you figure it out how to train the py-faster-rcnn using another dataset?

[aragon111]

@MarkoArsenovic I'm in the same situation...I found all the necessary steps for fast rcnn but still can't get any solution for py-faster-rcnn

[deboc]

I'm updating zeyuanxy's tutorial for py-faster-rcnn, I'll let you know

[aragon111]

@deboc that sounds great! Thank you!

[deboc]

I wasn't aware of the last updates with the config, that puzzled me a little ;) Here is my contribution, I hope it will help: https://github.com/deboc/py-faster-rcnn/blob/master/help/Readme.md

[deboc]

No you don't need to rename since the output dimension is not the same. Your issue looks really weird. Can you double check the 4 stage{something}train.pt files in your model folder ? The 84 shape in your error definitely seems to be an old 21x4...

[nazneenrajani]

Hi Can someone help me with running this ResNet model?

I modified the ResNet prototxt file to have the ROI proposals and was able to fine tune it successfully in 50K iterations on val set Train net output #3: rpn_loss_bbox = 0.0179207 (* 1 = 0.0179207 loss) Now I am trying to use the final caffe model along with the deploy.prototxt for ResNet modified as a test prototxt (removing lr params, top layer and appending the input data layer). However I keep getting this error. File "./lib/rpn/anchor_target_layer.py", line 116, in forward (all_anchors[:, 2] < im_info[0][1] ) & # width ValueError: operands could not be broadcast together with shapes (17100,) (600,800)

I printed a few debug statements but was unsuccessful in trouble shooting. Any help would be appreciated.

[Anubhav7]

Hi,

Has anyone tried to parallelize the calls to im_detect. I am facing issues with Caffe Initialization when I make calls to the network. Exact error details are below

Background: I have pre-loaded the network and am passing the net parameter in pooled calls.

Error Statement:

F0428 16:30:55.227715 19959 syncedmem.hpp:18] Check failed: error == cudaSuccess (3 vs. 0) initialization error

[leejiajun]

@Anubhav7 Out of memory？I guess.

[Anubhav7]

@leejiajun No, it's not one of those times :)

[anas-899]

Hi @deboc, Thank you for the very helpful training instructions in: https://github.com/deboc/py-faster-rcnn/blob/master/help/Readme.md Finally i successfully trained Inria dataset , But how i will do the test? I mean, I tried to use tools/test_net.py but it need test.prototxt file

[sahuvaibhav]

Hi, How did you do the refactoring got testset. it is not clear how to create symlinks for the testset. Please help

[BenjaminTT]

Hi all, Not sure that I am doing this the right way but it seems like the right places to ask some questions.

I am trying to get used to the training of the py-faster-rcnn and I have trained it with success on the B3DO dataset (though the results is only around 0.3 mAP) and my goal would be to perform object recognition using a Kinect on RGB an Depth separately. However, I have not come across articles dealing with object recognition using Depth only except some {bag of feature(HOG/ HDD) and classification} approaches. Is there any limitations I don't see that prevent considering using the faster RCNN on depth data only?

Also I tried to train the faster rcnn without using the pretrained models on imagenet. First on B3DO but success (I am assumed it was because of the size of the dataset), and then on pascal VOC (both end to end training and alt opt) but during the training part the loss does not seems to converge, and when I try to run the test_net.py I obtain an error whereas I could run it when I used the pretrained model : File "/home/benjamin/FastRCNN/py-faster-rcnn/tools/../lib/datasets/voc_eval.py", line 148, in voc_eval BB = BB[sorted_ind, :] IndexError: too many indices for array So if I would like not to use the pretrained model is there any steps aside not include --weights in the trainings options?

PS: I am using the ZF model If I am breaking any rules that I am unaware of, please tell me Any comment would be useful, thanks!

[tharuniitk]

Hi @deboc, Looking at the link below I followed the steps you have given for training faster RCNN using alternate optimization scheme on custom dataset. https://github.com/deboc/py-faster-rcnn/blob/master/help/Readme.md

I have 37 classes of data including background. I have changed the num_outputs accordingly. But I receive the following error again and again and can't find the mistake. I'll be extremely grateful if somebody can help me address this issue. im_proposals are calculated for all the training images and then an error is showing up as below in the attached image.

[deboc]

Hi, First are you sure to have removed the cache ? $ cd $ rm data/cache/voc_2016_train_gt_roidb.pkl $ rm output/faster_rcnn_alt_opt/voc_2016_train/vgg_*

[tharuniitk]

Hi @deboc I have cleared the cache. Now I have successfully trained the model. But I am meeting an error while testing same as that @BenjaminTT above. I am attaching it below. Grateful for your help.

Traceback (most recent call last): File "./tools/test_net.py", line 105, in test_net(net, imdb, max_per_image=args.max_per_image, vis=args.vis) File "/home/ashishkumar/py-faster-rcnn/tools/../lib/fast_rcnn/test.py", line 295, in test_net imdb.evaluate_detections(all_boxes, output_dir) File "/home/ashishkumar/py-faster-rcnn/tools/../lib/datasets/pascal_voc.py", line 341, in evaluate_detections self._do_python_eval(output_dir) File "/home/ashishkumar/py-faster-rcnn/tools/../lib/datasets/pascal_voc.py", line 304, in _do_python_eval use_07_metric=use_07_metric) File "/home/ashishkumar/py-faster-rcnn/tools/../lib/datasets/voc_eval.py", line 157, in voc_eval BB = BB[sorted_ind, :] IndexError: too many indices for array

[bityangke]

I also encountered the same problem as @tharuniitk

[tohnperfect]

Hi @deboc,

I had followed your guidance but I got the following error,

File "./tools/train_faster_rcnn_alt_opt.py", line 210, in <module> cfg_from_file(args.cfg_file) File "/home/caffe/py-faster-rcnn/tools/../lib/fast_rcnn/config.py", line 263, in cfg_from_file _merge_a_into_b(yaml_cfg, __C) File "/home/caffe/py-faster-rcnn/tools/../lib/fast_rcnn/config.py", line 235, in _merge_a_into_b raise KeyError('{} is not a valid config key'.format(k)) KeyError: 'MODEL_DIR is not a valid config key' I have no idea what did I do wrong. Could you please give me some advice?

[deboc]

Hi @tohnperfect, cfg.MODEL_DIR has been added in February, maybe you need to update your repo ?

[tohnperfect]

@deboc I just downloaded this git last week.

[deboc]

It's MODELS_DIR.

[deboc]

@tharuniitk, @bityangke: I think you're using the test step of voc dataset, but you need to write your own for the new dataset. For my tutorial on inria-person dataset the test step was not implemented, which I have done now. You can refer to this commit

[tohnperfect]

@deboc I've checked the codes and they are the lastest version

[deboc]

@tohnperfect Yes sorry. Like I said the parameter you are looking for is in fact "MODELS_DIR", not "MODEL_DIR". I think it's just that :)

[tohnperfect]

I see. Thank you very much @deboc.

[tohnperfect]

@deboc Are you planning to write a tutorial about testing on new datasets as well? I guess it would be a really useful tutorial.

[BenjaminTT]

hi @deboc, if the dataset is with the same format as pascal VOC the test implemented works, I trained on the B3DO dataset (with RGB only) and it worked (I obtain a mAP of ~0.32 ). However, problems appear when I am not using the pretrained model. In this case, whether I train on B3DO or pascal VOC, the test fail and even during training the loss does not seem to converge. Is there any specific step to do in order to avoid using the pretrained models ?

[bityangke]

@deboc Thank you very much. I am using the original VOC2007 datasets ，but I have changed the Annotations of trainval set（add more object boundings）.

[deboc]

@tohnperfect I dont think testing needs a complete tutorial but I can add some shell commands to do that in my existing training tutorial some day

@BenjaminTT I'm not sure why your test phase doesn't work, have you tried with a finetuned architecture like in my tuto (that is supposed to be more complicated with finetuning). For the complete training on pascal VOC... Well that's supposed to take some time! Have you considered to retrain the last layers only (setting the lower layers LR to 0) ? Debugging will be well faster then.

[BenjaminTT]

@deboc, thank you for your insight, I am using the end to end training so far and after some investigation the reason the test does not work is because the text file with the detection is empty. I am not sure but I think it may be because even if the training ran without error, it did not succeed and the net is unable to perform the detections, I will also try with your method.

The reason I am interested with performing the complete training is because I am considering on doing it on Depth maps and it that case the Imagenet Model would not be relevant to use. So do you have any advise for this? I guess I should increase the number of iterations but do I need to modify the learning rate or other values ?

Thank you again.

[tohnperfect]

What do I need to provide in test.mat when it is loaded by _selective_search_roidb? How the AP is affected if I set __C.TEST.HAS_RPN to True?

[fernandorovai]

@BenjaminTT, @tharuniitk Did you find out the problem? I am facing the same error "BB = BB[sorted_ind, :] IndexError: too many indices for array" after executing the sh script to train and test the network. I am using VGG_CNN_M_1024, end2end configuration and pretrained imagenet_model . Also, using the original VOC2007 dataset. I noticed that the losses are going up and down and never converge (70000 iterations). Anyone has any idea what is going on?

Thank you!

[tiepnh]

@fernandorovai : I have same problem with you. I just remove the initial weight in script file faster_rcnn_end2end.sh Any one know what happen?

[tiepnh]

@fernandorovai @BenjaminTT @tharuniitk : I just get help from @DiegoPortoJaccottet , he said that the training must use the pre-train network. If we do not use the pre-train the network, even the caffemodel is created, it still cannot perform detection/classification.

@deboc : in your tuto I see that you use the pre-train network, but if I create the new network by myself, there are any way to train the faster rcnn without pre-train network or how to create pre-train network for the new network?

[BenjaminTT]

Thank you @tiepnh, Does anyone knows if it is possible to find a way around? I still do not really understand why it is a necessity and not just something to speed up the learning

[LiberiFatali]

To train with your own network, I think you need to define the network in Caffe prototxt first, then train it on labeled datasets to get Caffe model. After that, you can use that model(trained by Caffe) in py-faster-rcnn. You also need to insert RPN layer into original prototxt to make it work in py-faster-rcnn.

[BenjaminTT]

Thank you very much for this informations @LiberiFatali. Can anyone can explain why the py-faster-rcnn can not be trained fully directly on the dataset that interest us? I do not see where is the constraint that impose to have a pretrained network. Tell me if I am wrong but if you do not precise the weights for the training aren't they generated with some functions for the initialization (randomly, or something else)? In that case the training should still be possible but more time consuming no?

[tiepnh]

Hi @BenjaminTT, not sure your found the solution or not. In case your still have no answer, check the item #238 to train the network without pretrained network

[raldam]

@leejiajun @LiberiFatali @banxiaduhuo

Hello, I encountered the same problem when training py-faster RCNN on imagenet data set. I am using this repository. In the dataset preparation code /tools/assemble_imagenet_train.py there is a section to filter out the BBOXes with inappropriate aspect ratio. Could you please give me the values you used for end2end training ?

[sounakdey]

In regards to the error i encountered i will be very happy if someone helps me out I followed this link https://github.com/deboc/py-faster-rcnn/blob/master/help/Readme.md and after executing this command $ ./tools/train_faster_rcnn_alt_opt.py --gpu 1 --net_name INRIA_Person --weights data/imagenet_models/VGG_CNN_M_1024.v2.caffemodel --imdb inria_train --cfg ./config.yml I had this error:

Output will be saved to /home/sounak/py-faster-rcnn/output/default/train Filtered 0 roidb entries: 1228 -> 1228 WARNING: Logging before InitGoogleLogging() is written to STDERR F0823 18:42:11.847455 4797 io.cpp:36] Check failed: fd != -1 (-1 vs. -1) File not found: $/home/sounak/py-faster-rcnn/models/INRIA_Person/faster_rcnn_alt_opt/stage1_rpn_solver60k80k.pt *** Check failure stack trace: ***

Please help

[useebear]

@sounakdey

1. check your folder for the file, it says File not found : $/home/sounak/py-faster-rcnn/models/INRIA_Person/faster_rcnn_alt_opt/stage1_rpn_solver60k80k.pt
2. The problem might be that you need to create the folder faster_rcnn_alt_opt/ under INRIA_Person, then copy all .pt files into it.
3. modify the path in the train_net path of the file stage1_rpn_solver60k80k.pt, and other similar .pt files

[duygusar]

@BenjaminTT @fernandorovai

I wonder if you guys have found a way out of this problem. I have the exact problem, I have made a few changes on ZF (end2end) net to work with RGB and depth - sort of a fusion - but the structure is basically the same. (I am actually testing it on RGB and RGB, I wanted to make sure it works before I work with depth images) And indeed it does not even converge on the training even after 70000 iterations and the problem is raised because there are no detections at testing stage. None for majority of the classes ( and very very very accuracy low for a few). I would be really happy if you can share your insights and experience with this problem. Thanks.

[BenjaminTT]

@duygusar Hi, it may be because the network is not actually learning on all the layers. You have to make sure that you initialize all of them (if you use the prototxt then some of them are not initialized because they use the weights of the pretrained models). If you have already done it and still have this problem try to increase a little the learning rate (since you are not using the pretrained model you sort of have more to learn). Hope this help.

[duygusar]

@BenjaminTT Thanks for the tip, I will try running with a higher learning rate. I have had recently tried initializing all convolution and innerproduct layers so I was able to make it work without any errors (the one caused during eval because of the empty detection sets) however I only have a a mAP of 0.25 even at 200.000 iterations.

[mbarkiwafa]

My problem that I haven't any evaluation :( ===> Index Error: too many indices . BB=BB[Sorted_ind,:]

I need help Thanks . ᐧ

2016-10-24 16:01 GMT+01:00 duygusar [email protected]:

@BenjaminTT https://github.com/BenjaminTT Thanks for the tip, I will try running with a higher learning rate. I have recently tried initializing all convolution and innerproduct layers so I was able to make it work without any errors (the one caused during eval because of the empty detection sets) however I only have a a mAP of 0.25 even at 200.000 iterations.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/rbgirshick/py-faster-rcnn/issues/27#issuecomment-255765945, or mute the thread https://github.com/notifications/unsubscribe-auth/ARcYFty1Az0IXVnRMBm_FQ_BgsXEnXbvks5q3Mg1gaJpZM4GmI20 .

[duygusar]

@mbarkiwafa It is most possible that your model is not converging either, the loss should decrease as the model trains, however it is not the case for me, it won't converge - no detections for one or more classes leads to that problem

[hengck23]

anyone tried training for two class using the inria dataset with "faster_rcnn_end2end.yml"?

[hawkjk]

Hello everyone， I try to train the Wider Face database on the py-faster-rcnn env. I had modified the database and finish the modification of relative files. Now the trainning can finished. But the mAP is always 0. Anyone know what happened? Thanks.

[slchiang]

@leejiajun @LiberiFatali @deboc Hi, I am training faster rcnn end-to-end with my own dataset. I always get core dump after a few iterations. I guess it is because the method _sample_rois in proposal_target_layer.py has no rois return since I printed out the array of rois and it is empty. Any suggestion that I can solve it or workaround?

In addition, I found that the gt_boxes seems different from the annotation I have, and seems flipped. I guess this is because I used __C.TRAIN.USE_FLIPPED = True in config.py. Does this matter?

Thank you in advance.

[LiberiFatali]

In my case, I modified https://github.com/rbgirshick/py-faster-rcnn/blob/master/lib/datasets/imdb.py#L218 and changed this line if gt_roidb is not None and gt_roidb[i]['boxes'].size > 0: to if gt_roidb[i] is not None and gt_roidb[i]['boxes'].size > 0:

[slchiang]

@LiberiFatali Thanks for your quick response. This seems not work for me, if _sample_rois in ProposalTargetLayer gives empty rois then the code will crash when it comes to top[0].reshape(*rois.shape) in the forward function of ProposalTargetLayer. Do you know where is the code calling ProposalTargetLayer or how can I deal with this case so that it skips current bottom data and fetch a new bottom data?

Thank you.

[franciszzj]

@leejiajun I changed my dataset,but I can train it on the original net at first,but when I changed the net, it can not work.The same dataset, but can not work on different net,I got this problem gt_argmax_overlaps = overlaps.argmax(axis=0) ValueError: attempt to get argmax of an empty sequence do u have any solvers?

[LiberiFatali]

@slchiang How about using

def filter_roidb(roidb):
    """Remove roidb entries that have no usable RoIs."""

at https://github.com/rbgirshick/py-faster-rcnn/blob/master/lib/fast_rcnn/train.py#L127 ?

[slchiang]

@LiberiFatali Thanks for your reply. But I guess this method is already called in train_net? Do you mean I need to call this method at somewhere else? So far, I started fine-tuning with low TRAIN.FG_THRESH and TRAIN.BG_THRESH_HI, and I increased the threshold gradually. Unfortunately, the loss does not decrease, and the result model is not good thought. Any suggestion for improving training results?

Thank you.

[hongkaichen]

@hawkjk hi, I also try to train my dataset on the py-faster-rcnn. The training can be finished. But the mAP is always 0. Do u have any solvers? Anyone know what happened? Thanks.

[SHEKOLDA]

@BenjaminTT @fernandorovai @duygusar I've encountered the same problem of IndexError: too many indices for array trying to train on my own dataset. Have you solved it? It's pretty wierd and seemingly has something to do with how it writes the results. detections.pkl isn't empty but all of the results txt files are. The net also doesn't converge but I guess it's more of a finetuning problem rather then configuration (although I may be wrong on that).

[dantp-ai]

This is not quite true @yileo19920925, the demo from @deboc assumes that your annotations files have .txt format. This can be verified by looking at the method _load_inria_annotation documentation, namely

Load image and bounding boxes info from txt files of INRIAPerson.

@leefionglee If your application uses .xml formatted annotations file, then take a look at pascal_voc.py and voc_eval.py

[xinleipan]

try this: https://github.com/xinleipan/py-faster-rcnn-with-new-dataset

[hanjf12]

@Franciszzj I met the same error to you, do you have solved it?

[vabrishami]

@BenjaminTT Did you solve your problem by initializing the convolution and mult layers with random weights?

[BenjaminTT]

@vabrishami Yes for me the solution was to make sure all the weight were initialized and then the training would work. It even worked on a dataset of depth map that were registered as png (so all the three channels were the same)

[vabrishami]

@BenjaminTT Is it possible for you to share your train.prototxt?

[BenjaminTT]

@vabrishami Here it is, I hope it will help you (this is for 19 classes + background). train.txt

[vabrishami]

@BenjaminTT Thanks Benjamin. It was really useful for me. I have another question. Why you used three same channel? Why not a single channel image?

[BenjaminTT]

@vabrishami It is just a matter of convenience, I am using 2 networks trained independently, one on colour and one on Depth so it enables me to use the same file and also it enables to perform the training without preprocessing the images from the SUNRGBD dataset

[indsak]

Hello everyone i am trying to train faster RCNN as per https://github.com/deboc/py-faster-rcnn/tree/master/help

When i try to train using the command ./tools/train_faster_rcnn_alt_opt.py --gpu 0 --net_name fishclassify --weights data/imagenet_models/VGG_CNN_M_1024.v2.caffemodel --imdb fishclassify --cfg /home/fast-rcnn/py-faster-rcnn/config.yml

I am getting the following error File "/fast-rcnn/py-faster-rcnn/tools/../lib/datasets/factory.py", line 46, in get_imdb raise KeyError('Unknown dataset: {}'.format(name))

Any help on what might be causing this? I had created fishclassify.py and fishclassify_eval.py under lib/datasets as well as modified factory.py

When i open the .py files in pycharm it shows unresolved import for "import datasets". Any idea where i had gone wrong?

[deboc]

Hi, according to my tutorial

--imdb is the full name of the database as specified in the lib/datasets/factory.py file (nb: dont forget to add the test/train suffix !)

So in your case

--imdb fishclassify_train

[indsak]

Thank you very much. Yesterday i had tried that..may be because of some other problem i was getting the same error. Now i tried your suggestion and the error is gone.But some AssertionError for the paths of the files. I will clear that. Thank you very much

[indsak]

@deboc Hi, One more doubt, does the code strip the .xml extension in the newdataset.py file? When i run the code AssertionError coming for the image file because it searches for file.xml.jpg which is not present.

[deboc]

Sorry but which .xml extension ? In my example of INRIA person dataset, I list .txt annotation files and write a train.txt index file with image names without extension (they are stripped with a sed command line). You should adapt that to your new dataset.

[indsak]

@deboc ok.Thank you. I had written the train.txt with .xml extension. That was creating the issue. one more doubt. The name given the annotation file and those declared in self._classes should match right? I had written the abbreviated form in self._classes and the full form was retrieved from annotation files and it showed error. I hope both should match. Sorry to bother with silly doubts.

[indsak]

@deboc Can you help me pls I am getting error at cls = self._class_to_ind[obj.find('name').text.lower().strip()]

in the newdataset.py file

In the self._classes i had declared the names as per the names given in the Annotation file. Its not entirely in small letters. Is this the problem?

[deboc]

Please study carefully the tutorial applied to INRIA dataset before trying to use your dataset.

The method _class_to_ind required your annotations to match the index of the classes list (1 for pedestrian in the tutorial, forget about 0 it is internally used for background). So depending on your classes list you should have integers as annotations. You can also hack the newdataset.py instead, to match your annotations cls = obj.find('name').text.lower().strip() Should do the trick

[indsak]

@deboc I successfully started training. But got the following error

line 118, in rpn_roidb if int(self._year) == 2007 or self._image_set != 'test': AttributeError: 'fishclassify' object has no attribute '_year'

Any help on how to solve this?

[indsak]

Thank you for your help.

On Jun 13, 2017 8:19 PM, "Pierre de Beaucorps" [email protected] wrote:

Please study carefully the tutorial applied to INRIA dataset before trying to use your dataset.

The method _class_to_ind required your annotations to match the index of the classes list (1 for pedestrian in the tutorial, forget about 0 it is internally used for background). So depending on your classes list you should have integers as annotations. You can also hack the newdataset.py instead, to match your annotations cls = obj.find('name').text.lower().strip() Should do the trick

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/rbgirshick/py-faster-rcnn/issues/27#issuecomment-308141246, or mute the thread https://github.com/notifications/unsubscribe-auth/AWAQiRZbMOyLWMsG0_41SyLgXiQoNDv_ks5sDqFhgaJpZM4GmI20 .

[indsak]

@deboc Any help on the below problem

line 118, in rpn_roidb if int(self._year) == 2007 or self._image_set != 'test': AttributeError: 'fishclassify' object has no attribute '_year'

Any help on how to solve this?

[zhenni]

@indsak Check your script TRAIN_IMDB="voc_2007_trainval" you need a year like 2007. Or you can modify the code in lib/datasets/factory.py when you register your own dataset and edit all parts with year in your own dataset reader like lib/datasets/pascal_voc.py