Open kkhuang1990 opened 7 years ago
Yes the external dependencies are a bit tricky, or can be @kkhuang1990 :S
I think it would be interesting to try the dataset, and the same basic architecture with that DeepMedic for example: https://github.com/Kamnitsask/deepmedic
Or convert the simple 2D+3D into a residual architecture? And as Theano gets discontinued, maybe implement the whole thing from scratch using Tensorflow or PyTorch?
Or just having to redo the configs and all based on the latest code from ZNN guys if you want more memory without the GPU acceleration? https://github.com/seung-lab/znn-release
when I run the command below to train the model python train.py -c ../../configs/ZNN_configs/config_VD2D_tanh.cfg error like this came out Traceback (most recent call last): File "train.py", line 263, in
main( args )
File "train.py", line 77, in main
smp_trn = zsample.CSamples(config, pars, pars['train_range'], net, outsz, logfile)
File "/home/mil/huang/Coronary_Artery_Segmentation/vesselNN/znn-release/python/front_end/zsample.py", line 551, in init
sample = CBoundarySample(config, pars, sid, net, outsz, log)
File "/home/mil/huang/Coronary_Artery_Segmentation/vesselNN/znn-release/python/front_end/zsample.py", line 422, in init
log=log, is_forward=is_forward)
File "/home/mil/huang/Coronary_Artery_Segmentation/vesselNN/znn-release/python/front_end/zsample.py", line 58, in init
outsz, setsz_in, fov, is_forward=is_forward )
File "/home/mil/huang/Coronary_Artery_Segmentation/vesselNN/znn-release/python/front_end/zdataset.py", line 377, in init
outsz, setsz, mapsz, is_forward=is_forward )
File "/home/mil/huang/Coronary_Artery_Segmentation/vesselNN/znn-release/python/front_end/zdataset.py", line 229, in init
arrlist = self._read_files( fnames );
File "/home/mil/huang/Coronary_Artery_Segmentation/vesselNN/znn-release/python/front_end/zdataset.py", line 305, in _read_files
vol = emirt.emio.imread(fl)
AttributeError: 'module' object has no attribute 'emio'
I have already downloaded the latest emirt from github and put it under the znn-release/python dir