CNTK objectdetection sample 4_trainSVM.py not running

[netvinod]

I am trying to build and run the ObjectDetectionUsingCNTK using grocery data set.

I am running into issues while running it.
The 1_, 2_ and 3_ files compile fine. When i attempt to compile 4_TrainSVM, i run into the issue shown below:
Traceback (most recent call last):
File "4_trainSvm.py", line 50, in
svm_targetNorm, svm_penality, svm_loss, svm_rngSeed)
File "c:\Repo\ml_dl_practice\object_detection_cntk_svm\ObjectDetectionUsingCnt
k\fastRCNN\train_svms.py", line 45, in __init__
self.feature_scale = self._get_feature_scale()
File "c:\Repo\ml_dl_practice\object_detection_cntk_svm\ObjectDetectionUsingCnt
k\fastRCNN\train_svms.py", line 74, in _get_feature_scale
total_norm / count, total_sum / count))
TypeError: unsupported format string passed to numpy.ndarray.__format__

Any pointers to fixing this issue? I see this issue both in python3.6 and python 3.5. If I comment the above line (its a print statement) it proceeds, but further down the line I run into issue shown below:

Traceback (most recent call last):
File "4_trainSvm.py", line 51, in
svm.train()
File "c:\Repo\ml_dl_practice\object_detection_cntk_svm\ObjectDetectionUsingCnt
k\fastRCNN\train_svms.py", line 187, in train
self.get_pos_examples()
File "c:\Repo\ml_dl_practice\object_detection_cntk_svm\ObjectDetectionUsingCnt
k\fastRCNN\train_svms.py", line 107, in get_pos_examples
scores, boxes, feat = self.im_detect(self.net, i, gt_boxes, self.feature_sca
le, gt_inds, boReturnClassifierScore = False)
File "c:\Repo\ml_dl_practice\object_detection_cntk_svm\ObjectDetectionUsingCnt
k\helpers.py", line 620, in im_detect
if bboxIndices != None:
ValueError: The truth value of an array with more than one element is ambiguous.
Use a.any() or a.all()

My numpy is 1.13 and running on a conda python installation on Windows 8

[kpierce8]

I don't know if it is right but I changed the line to
if np.any(bboxIndices != None):
And it ran.
Well it found a new error in scikit-learn which I'm now trying to track down.

[brunoprojects]

Same here, changed to np.any(), but now stuck with sklearn issue. Any progress?

[kpierce8]

Ah, yes. Change line 69 in the parameters file to "balanced" from "auto"