KeyError: 'fc7_conv'

[albertyou2]

hi @jinfagang

when I run 'python train_ssd_kitti.py', I got this error :

Traceback (most recent call last):
File "train_ssd_kitti.py", line 455, in
prior_variance=prior_variance, kernel_size=3, pad=1, lr_mult=lr_mult)
File "/home/sojoyoo/caffe_ssd/python/caffe/model_libs.py", line 870, in CreateMultiBoxHead
num_output=num_loc_output, kernel_size=kernel_size, pad=pad, stride=1, **bn_param)
File "/home/sojoyoo/caffe_ssd/python/caffe/model_libs.py", line 97, in ConvBNLayer
net[conv_name] = L.Convolution(net[from_layer], num_output=num_output,
File "/home/sojoyoo/caffe_ssd/python/caffe/net_spec.py", line 188, in getitem
return self.getattr(item)
File "/home/sojoyoo/caffe_ssd/python/caffe/net_spec.py", line 182, in getattr
return self.tops[name]
KeyError: 'fc7_conv'

Have you met this before?

[wenchima]

hi @albertyou2

Have you solved your problem? I changed the fc7_conv into fc7 and it worked well. But my problem is that my training loss is always be zero. Have you kept doing on this?

[albertyou2]

@wenchima
No , I didn't solve this.
I 'm now switch to caffe_ssd .

[nattari]

Facing the same issue. Anyone solved it yet?

[ss32]

Same error, no luck

Looking at https://github.com/weiliu89/caffe/blob/ssd/examples/ssd/ssd_pascal.py and comparing, they use fc7 instead of fc7_conv. Make the change and the script will run without error