need more than 1 value to unpack

Hey guys, when I run inference with ipyn (VOCpascal) I get this error:

`--------------------------------------------------------------------------- ValueError Traceback (most recent call last) <ipython-input-6-f56c15ba16bb> in <module>() ----> 1 image_batch, outputs = next(test_generator) 2 image = image_batch[0, …].copy() 3 image -= min(image.flatten()) 4 image /= max(image.flatten()) 5 image *= 255

/usr/local/lib/python2.7/dist-packages/keras_retinanet/preprocessing/pascal_voc.pyc in next(self) 179 # lock indexing to prevent race conditions 180 with self.lock: –> 181 selection, _, batch_size = next(self.index_generator) 182 183 result = self._get_batches_of_transformed_samples(selection)

ValueError: need more than 1 value to unpack`

the initialization of data throws no error, actually gives no feedback at all, but I think its fine. path should be correct, test.txt exists and contains just one line: the name of the image (without .jpg)

I saw here that someone had kind of the same error. downgrading to keras 2.0.8 throws a new error at a prior step, so I would like to avoid that. I think I have all the changes from the PR mentioned in the other issue.

Any ideas or advice or workarounds?

versions: python 2.7.6 keras 2.0.9

[nealzGu]

ahhhh right, that thing. thanks again, everything is working fine now. you helped me a lot 😃

[hgaiser]

By default python will use the version from your /usr/ directory. You can run your programs with PYTHONPATH=/path/to/your/cloned/keras-retinanet python examples/train_pascal.py. This will make python have a preference of using your local (non-installed) copy.

unrelated: it is working now (at least not throwing errors). I trained outside of the ipyn and want to run inference now. If I wanna use the ipyn, I just skip the “train on data” part, right? But don’t I skip the “load pretrained weights” part in this case as well?! Sorry, different topic, I will close this thread directly afterwards 😃

Yeah exactly. If you change the weights parameter when creating the model to point to your saved model, it should use those weights. ie.:

model = keras_retinanet.models.ResNet50RetinaNet(image_input, num_classes=20, nms=True, weights='snapshots/resnet50_voc_best.h5')