Closed kilickaya closed 8 years ago
Hmm, we did all the conversions on Linux machines :( But I'm surprised that there is such a limit and I was not able to find any further information, only similar thing was here which suggests to use newer version of numpy...
Hey @kilickaya Did you manage to convert the Places database to MatConvNet? Which version of the pertained cnn did you use? Alexnet or VGG16?
I manage to convert Places (along with HYBRID etc.) not by Windows (due to Python Memory issues) but by Linux
@kilickaya Do you mind sharing the files used to convert the places-database to Matconvnet? It would be really helpful!
First, places is not a database, it is a CNN architecture (sharing the same architecture with famous AlexNet) where MIT researchers trained to recognize scenes -or places in paper's jargon- (instead of ImageNet objects).
Then, Yes I can share these models (along with the HYBRID model) via DropBox or Google Drive, will update this when upload is complete.
Thank you! Let me know when you have the link :) (I meant a 'CNN trained on the places database', but you are absolutely right that I wrote it incorrect)
Hi @kilickaya , Could you please share the models in Matconvnet format?
Thanks,
I have been trying to convert the Places 205 VGG CNN model, everything runs fine, but the weight matrices are empty. Dont know where is it going wrong. Can you please share your converted models, and possibly the scripts.
Hi guys, I'm also looking for the Places 205 VGG CNN model in matconvnet format can someone share it please? Thanks a lot.
Hi, @kilickaya Could you please share the Places models in Matconvnet format? Thank you so much!
hi,guys, I am a newer in python and caffe, I want to convert a pre-trained caffemodel to matconvnet format, does anyone could tell me how to use the import-caffe.py? what are the inputs? Thanks a lot
Hi,
I am running MatConvNet on a 64 bit Windows machine.
I am using import-caffe.py to convert a proposed model (http://places.csail.mit.edu/) to MatConvNet format. It succesfully converts the model until the 15th layer.
However, after that, I receive a 'Memory Error'. I suspect this is due to the fact that Windows only allows 1 GB RAM usage for a single (python) process.
I would like to ask if there is any workaround exists for such error?