Python numpy.squeeze() Examples

def draw_heatmap(img, heatmap, alpha=0.5):
    """Draw a heatmap overlay over an image."""
    assert len(heatmap.shape) == 2 or \
        (len(heatmap.shape) == 3 and heatmap.shape[2] == 1)
    assert img.dtype in [np.uint8, np.int32, np.int64]
    assert heatmap.dtype in [np.float32, np.float64]

    if img.shape[0:2] != heatmap.shape[0:2]:
        heatmap_rs = np.clip(heatmap * 255, 0, 255).astype(np.uint8)
        heatmap_rs = ia.imresize_single_image(
            heatmap_rs[..., np.newaxis],
            img.shape[0:2],
            interpolation="nearest"
        )
        heatmap = np.squeeze(heatmap_rs) / 255.0

    cmap = plt.get_cmap('jet')
    heatmap_cmapped = cmap(heatmap)
    heatmap_cmapped = np.delete(heatmap_cmapped, 3, 2)
    heatmap_cmapped = heatmap_cmapped * 255
    mix = (1-alpha) * img + alpha * heatmap_cmapped
    mix = np.clip(mix, 0, 255).astype(np.uint8)
    return mix 

def cleverhans_attack_wrapper(cleverhans_attack_fn, reset=True):
    def attack(a):
        session = tf.Session()
        with session.as_default():
            model = RVBCleverhansModel(a)
            adversarial_image = cleverhans_attack_fn(model, session, a)
            adversarial_image = np.squeeze(adversarial_image, axis=0)
            if reset:
                # optionally, reset to ignore other adversarials
                # found during the search
                a._reset()
            # run predictions to make sure the returned adversarial
            # is taken into account
            min_, max_ = a.bounds()
            adversarial_image = np.clip(adversarial_image, min_, max_)
            a.predictions(adversarial_image)
    return attack 

def __getitem__(self, idx):
        if self.fasta_extractor is None:
            self.fasta_extractor = FastaExtractor(self.fasta_file)
        interval = self.bt[idx]

        if interval.stop - interval.start != self.SEQ_WIDTH:
            raise ValueError("Expected the interval to be {0} wide. Recieved stop - start = {1}".
                             format(self.SEQ_WIDTH, interval.stop - interval.start))

        # Run the fasta extractor
        seq = np.squeeze(self.fasta_extractor([interval]), axis=0)
        return {
            "inputs": {"dna": seq},
            "metadata": {
                "ranges": GenomicRanges.from_interval(interval)
            }
        } 

def __getitem__(self, idx):
        if self.fasta_extractor is None:
            self.fasta_extractor = FastaExtractor(self.fasta_file)
        interval = self.bt[idx]

        if interval.stop - interval.start != self.SEQ_WIDTH:
            raise ValueError("Expected the interval to be {0} wide. Recieved stop - start = {1}".
                             format(self.SEQ_WIDTH, interval.stop - interval.start))

        # Run the fasta extractor
        seq = np.squeeze(self.fasta_extractor([interval]), axis=0)
        return {
            "inputs": {"dna": seq},
            "metadata": {
                "ranges": GenomicRanges.from_interval(interval)
            }
        } 

def __getitem__(self, idx):
        if self.fasta_extractor is None:
            self.fasta_extractor = FastaExtractor(self.fasta_file)
        interval = self.bt[idx]

        if interval.stop - interval.start != self.SEQ_WIDTH:
            raise ValueError("Expected the interval to be {0} wide. Recieved stop - start = {1}".
                             format(self.SEQ_WIDTH, interval.stop - interval.start))

        # Run the fasta extractor
        seq = np.squeeze(self.fasta_extractor([interval]), axis=0)
        return {
            "inputs": {"dna": seq},
            "metadata": {
                "ranges": GenomicRanges.from_interval(interval)
            }
        } 

def __getitem__(self, idx):
        if self.fasta_extractor is None:
            self.fasta_extractor = FastaExtractor(self.fasta_file)
        interval = self.bt[idx]

        if interval.stop - interval.start != self.SEQ_WIDTH:
            raise ValueError("Expected the interval to be {0} wide. Recieved stop - start = {1}".
                             format(self.SEQ_WIDTH, interval.stop - interval.start))

        # Run the fasta extractor
        seq = np.squeeze(self.fasta_extractor([interval]), axis=0)
        return {
            "inputs": {"dna": seq},
            "metadata": {
                "ranges": GenomicRanges.from_interval(interval)
            }
        } 

def __getitem__(self, idx):
        if self.fasta_extractor is None:
            self.fasta_extractor = FastaExtractor(self.fasta_file)
        interval = self.bt[idx]

        if interval.stop - interval.start != self.SEQ_WIDTH:
            raise ValueError("Expected the interval to be {0} wide. Recieved stop - start = {1}".
                             format(self.SEQ_WIDTH, interval.stop - interval.start))

        # Run the fasta extractor
        seq = np.squeeze(self.fasta_extractor([interval]), axis=0)
        return {
            "inputs": {"dna": seq},
            "metadata": {
                "ranges": GenomicRanges.from_interval(interval)
            }
        } 

def apply_affine(aff, coords):
    '''
    apply_affine(affine, coords) yields the result of applying the given affine transformation to
      the given coordinate or coordinates.

    This function expects coords to be a (dims X n) matrix but if the first dimension is neither 2
    nor 3, coords.T is used; i.e.:
      apply_affine(affine3x3, coords2xN) ==> newcoords2xN
      apply_affine(affine4x4, coords3xN) ==> newcoords3xN
      apply_affine(affine3x3, coordsNx2) ==> newcoordsNx2 (for N != 2)
      apply_affine(affine4x4, coordsNx3) ==> newcoordsNx3 (for N != 3)
    '''
    if aff is None: return coords
    (coords,tr) = (np.asanyarray(coords), False)
    if len(coords.shape) == 1: return np.squeeze(apply_affine(np.reshape(coords, (-1,1)), aff))
    elif len(coords.shape) > 2: raise ValueError('cannot apply affine to ND-array for N > 2')
    if   len(coords) == 2: aff = to_affine(aff, 2)
    elif len(coords) == 3: aff = to_affine(aff, 3)
    else: (coords,aff,tr) = (coords.T, to_affine(aff, coords.shape[1]), True)
    r = np.dot(aff, np.vstack([coords, np.ones([1,coords.shape[1]])]))[:-1]
    return r.T if tr else r 

def visualize(net, preprocessed_img, orig_img, conv_layer_name):
    # Returns grad-cam heatmap, guided grad-cam, guided grad-cam saliency
    imggrad = get_image_grad(net, preprocessed_img)
    conv_out, conv_out_grad = get_conv_out_grad(net, preprocessed_img, conv_layer_name=conv_layer_name)

    cam = get_cam(imggrad, conv_out)
    
    ggcam = get_guided_grad_cam(cam, imggrad)
    img_ggcam = grad_to_image(ggcam)
    
    img_heatmap = get_img_heatmap(orig_img, cam)
    
    ggcam_gray = to_grayscale(ggcam)
    img_ggcam_gray = np.squeeze(grad_to_image(ggcam_gray))
    
    return img_heatmap, img_ggcam, img_ggcam_gray 

def decode_topk(self, sess, latest_tokens, enc_top_states, dec_init_states):
    """Return the topK results and new decoder states."""
    feed = {
        self._enc_top_states: enc_top_states,
        self._dec_in_state:
            np.squeeze(np.array(dec_init_states)),
        self._abstracts:
            np.transpose(np.array([latest_tokens])),
        self._abstract_lens: np.ones([len(dec_init_states)], np.int32)}

    results = sess.run(
        [self._topk_ids, self._topk_log_probs, self._dec_out_state],
        feed_dict=feed)

    ids, probs, states = results[0], results[1], results[2]
    new_states = [s for s in states]
    return ids, probs, new_states 

def predict(self, batch_inputs, batch_ruitu):
        assert batch_ruitu.shape[0] == batch_inputs.shape[0], 'Shape Error'
        assert batch_inputs.shape[1] == 28 and batch_inputs.shape[2] == 10 and batch_inputs.shape[3] == 9, 'Error! Obs input shape must be (None, 28,10,9)'
        assert batch_ruitu.shape[1] == 37 and batch_ruitu.shape[2] == 10 and batch_ruitu.shape[3] == 29, 'Error! Ruitu input shape must be (None, 37,10, 29)'
        #all_pred={}
        pred_result_list = []
        for i in range(10):
            #print('Predict for station: 9000{}'.format(i+1))
            result = self.model.predict(x=[batch_inputs[:,:,i,:], batch_ruitu[:,:,i,:]])
            result = np.squeeze(result, axis=0)
            #all_pred[i] = result
            pred_result_list.append(result)
            #pass

        pred_result = np.stack(pred_result_list, axis=0)
        #return all_pred, pred_result
        print('Predict shape (10,37,3) means (stationID, timestep, features). Features include: t2m, rh2m and w10m')
        self.pred_result = pred_result
        return pred_result 

def predict(self, batch_inputs, batch_ruitu, batch_ids):
        #assert batch_ruitu.shape[0] == batch_inputs.shape[0], 'Shape Error'
        #assert batch_inputs.shape[1] == 28 and batch_inputs.shape[2] == 10 and batch_inputs.shape[3] == 9, 'Error! Obs input shape must be (None, 28,10,9)'
        #assert batch_ruitu.shape[1] == 37 and batch_ruitu.shape[2] == 10 and batch_ruitu.shape[3] == 29, 'Error! Ruitu input shape must be (None, 37,10, 29)'
        pred_result_list = []
        for i in range(10):
            #print('Predict for station: 9000{}'.format(i+1))
            result = self.model.predict(x=[batch_inputs[:,:,i,:], batch_ruitu[:,:,i,:], batch_ids[:,:,i]])
            result = np.squeeze(result, axis=0)
            #all_pred[i] = result
            pred_result_list.append(result)
            #pass

        pred_result = np.stack(pred_result_list, axis=0)
        #return all_pred, pred_result
        print('Predict shape (10,37,3) means (stationID, timestep, features). Features include: t2m, rh2m and w10m')
        self.pred_result = pred_result
        return pred_result 

def set_recall(predictions, labels, weights_fn=common_layers.weights_nonzero):
  """Recall of set predictions.

  Args:
    predictions : A Tensor of scores of shape [batch, nlabels].
    labels: A Tensor of int32s giving true set elements,
      of shape [batch, seq_length].
    weights_fn: A function to weight the elements.

  Returns:
    hits: A Tensor of shape [batch, nlabels].
    weights: A Tensor of shape [batch, nlabels].
  """
  with tf.variable_scope("set_recall", values=[predictions, labels]):
    labels = tf.squeeze(labels, [2, 3])
    weights = weights_fn(labels)
    labels = tf.one_hot(labels, predictions.shape[-1])
    labels = tf.reduce_max(labels, axis=1)
    labels = tf.cast(labels, tf.bool)
    return tf.to_float(tf.equal(labels, predictions)), weights 

def __init__(self, x0, mu, epsmult = 4.0, noc = False):
        #determine number of planets and validate input
        nplanets = x0.size/6.
        if (nplanets - np.floor(nplanets) > 0):
            raise Exception('The length of x0 must be a multiple of 6.')
        
        if (mu.size != nplanets):
            raise Exception('The length of mu must be the length of x0 divided by 6')
        
        self.nplanets = int(nplanets)
        self.mu = np.squeeze(mu)
        if (self.mu.size == 1):
            self.mu = np.array(mu)
        
        self.epsmult = epsmult
        
        if not(noc) and ('EXOSIMS.util.KeplerSTM_C.CyKeplerSTM' in sys.modules):
            self.havec = True
            self.x0 = np.squeeze(x0)
        else:
            self.havec = False
            self.updateState(np.squeeze(x0)) 

def _heatmap_to_rects(self, grid_pred, bb_img):
        """Convert a heatmap to rectangles / bounding box candidates."""
        grid_pred = np.squeeze(grid_pred) # (1, H, W) => (H, W)

        # remove low activations
        grid_thresh = grid_pred >= self.heatmap_activation_threshold

        # find connected components
        grid_labeled, num_labels = morphology.label(
            grid_thresh, background=0, connectivity=1, return_num=True
        )

        # for each connected components,
        # - draw a bounding box around it,
        # - shrink the bounding box to optimal size
        # - estimate a score/confidence value
        bbs = []
        for label in range(1, num_labels+1):
            (yy, xx) = np.nonzero(grid_labeled == label)
            min_y, max_y = np.min(yy), np.max(yy)
            min_x, max_x = np.min(xx), np.max(xx)
            rect = RectangleOnImage(x1=min_x, x2=max_x+1, y1=min_y, y2=max_y+1, shape=grid_labeled)
            activation = self._rect_to_score(rect, grid_pred)
            rect_shrunk, activation_shrunk = self._shrink(grid_pred, rect)
            rect_rs_shrunk = rect_shrunk.on(bb_img)
            bbs.append((rect_rs_shrunk, activation_shrunk))

        return bbs 

def generate_video_image(batch_idx, examples, model):
    """Generate frames for a video of the training progress.
    Each frame contains N examples shown in a grid. Each example shows
    the input image and the main heatmap predicted by the model."""
    start_time = time.time()
    #print("A", time.time() - start_time)
    model.eval()

    # fw through network
    inputs, outputs_gt = examples_to_batch(examples, iaa.Noop())
    inputs_torch = torch.from_numpy(inputs)
    inputs_torch = Variable(inputs_torch, volatile=True)
    if GPU >= 0:
        inputs_torch = inputs_torch.cuda(GPU)
    outputs_pred_torch = model(inputs_torch)
    #print("B", time.time() - start_time)

    outputs_pred = outputs_pred_torch.cpu().data.numpy()
    inputs = (inputs * 255).astype(np.uint8).transpose(0, 2, 3, 1)
    #print("C", time.time() - start_time)
    heatmaps = []
    for i in range(inputs.shape[0]):
        hm_drawn = draw_heatmap(inputs[i], np.squeeze(outputs_pred[i][0]), alpha=0.5)
        heatmaps.append(hm_drawn)
    #print("D", time.time() - start_time)
    grid = ia.draw_grid(heatmaps, cols=11, rows=6).astype(np.uint8)
    #grid_rs = misc.imresize(grid, (720-32, 1280-32))
    # pad by 42 for the text and to get the image to 720p aspect ratio
    grid_pad = np.pad(grid, ((0, 42), (0, 0), (0, 0)), mode="constant")
    grid_pad_text = ia.draw_text(
        grid_pad,
        x=grid_pad.shape[1]-220,
        y=grid_pad.shape[0]-35,
        text="Batch %05d" % (batch_idx,),
        color=[255, 255, 255]
    )
    #print("E", time.time() - start_time)
    return grid_pad_text 

def Rmtx_ri(coef_ri, K, D, L):
    coef_ri = np.squeeze(coef_ri)
    coef_r = coef_ri[:K + 1]
    coef_i = coef_ri[K + 1:]
    R_r = linalg.toeplitz(np.concatenate((np.array([coef_r[-1]]),
                                          np.zeros(L - K - 1))),
                          np.concatenate((coef_r[::-1],
                                          np.zeros(L - K - 1)))
                          )
    R_i = linalg.toeplitz(np.concatenate((np.array([coef_i[-1]]),
                                          np.zeros(L - K - 1))),
                          np.concatenate((coef_i[::-1],
                                          np.zeros(L - K - 1)))
                          )
    return np.dot(np.vstack((np.hstack((R_r, -R_i)), np.hstack((R_i, R_r)))), D) 

def _logits_op(self, x, name=None):
        with tf.name_scope(name, "logits", [x]) as name:

            num_classes = self.adversarial.num_classes()

            def _backward_py(gradient_y, x):
                x = np.squeeze(x, axis=0)
                gradient_y = np.squeeze(gradient_y, axis=0)
                gradient_x = self.adversarial.backward(gradient_y, x)
                gradient_x = gradient_x.astype(np.float32)
                return gradient_x[np.newaxis]

            def _backward_tf(op, grad):
                images = op.inputs[0]
                gradient_x = tf.py_func(
                    _backward_py, [grad, images], tf.float32)
                gradient_x.set_shape(images.shape)
                return gradient_x

            def _forward_py(x):
                predictions = self.adversarial.batch_predictions(
                    x, strict=False)[0]
                predictions = predictions.astype(np.float32)
                return predictions

            op = py_func_grad(
                _forward_py,
                [x],
                [tf.float32],
                name=name,
                grad=_backward_tf)

            logits = op[0]
            logits.set_shape((x.shape[0], num_classes))

        return logits 

def pair_visual(original, adversarial, figure=None):
    """
    This function displays two images: the original and the adversarial sample
    :param original: the original input
    :param adversarial: the input after perterbations have been applied
    :param figure: if we've already displayed images, use the same plot
    :return: the matplot figure to reuse for future samples
    """
    import matplotlib.pyplot as plt

    # Squeeze the image to remove single-dimensional entries from array shape
    original = np.squeeze(original)
    adversarial = np.squeeze(adversarial)

    # Ensure our inputs are of proper shape
    assert(len(original.shape) == 2 or len(original.shape) == 3)

    # To avoid creating figures per input sample, reuse the sample plot
    if figure is None:
        plt.ion()
        figure = plt.figure()
        figure.canvas.set_window_title('Cleverhans: Pair Visualization')

    # Add the images to the plot
    perterbations = adversarial - original
    for index, image in enumerate((original, perterbations, adversarial)):
        figure.add_subplot(1, 3, index + 1)
        plt.axis('off')

        # If the image is 2D, then we have 1 color channel
        if len(image.shape) == 2:
            plt.imshow(image, cmap='gray')
        else:
            plt.imshow(image)

        # Give the plot some time to update
        plt.pause(0.01)

    # Draw the plot and return
    plt.show()
    return figure 

def __getitem__(self, idx):
        if self.seq_extractor is None:
            self.seq_extractor = FastaExtractor(self.fasta_file)
            self.dist_extractor = DistToClosestLandmarkExtractor(gtf_file=self.gtf,
                                                                 landmarks=ALL_LANDMARKS)

        interval = self.bt[idx]

        if interval.stop - interval.start != self.SEQ_WIDTH:
            raise ValueError("Expected the interval to be {0} wide. Recieved stop - start = {1}".
                             format(self.SEQ_WIDTH, interval.stop - interval.start))
        out = {}
        out['inputs'] = {}
        # input - sequence
        out['inputs']['seq'] = np.squeeze(self.seq_extractor([interval]), axis=0)

        # input - distance
        dist_dict = self.dist_transformer.transform(self.dist_extractor([interval]))
        dist_dict = {k: np.squeeze(v, axis=0) for k, v in dist_dict.items()}  # squeeze the batch axis
        out['inputs'] = {**out['inputs'], **dist_dict}

        # targets
        if self.target_dataset is not None:
            out["targets"] = np.array([self.target_dataset[idx]])

        # metadata
        out['metadata'] = {}
        out['metadata']['ranges'] = GenomicRanges.from_interval(interval)

        return out 

def extract_vggish_features(paths, path2gt, model): 
    """Extracts VGGish features and their corresponding ground_truth and identifiers (the path).

       VGGish features are extracted from non-overlapping audio patches of 0.96 seconds, 
       where each audio patch covers 64 mel bands and 96 frames of 10 ms each.

       We repeat ground_truth and identifiers to fit the number of extracted VGGish features.
    """
    # 1) Extract log-mel spectrograms
    first_audio = True
    for p in paths:
        if first_audio:
            input_data = vggish_input.wavfile_to_examples(config['audio_folder'] + p)
            ground_truth = np.repeat(path2gt[p], input_data.shape[0], axis=0)
            identifiers = np.repeat(p, input_data.shape[0], axis=0)
            first_audio = False
        else:
            tmp_in = vggish_input.wavfile_to_examples(config['audio_folder'] + p)
            input_data = np.concatenate((input_data, tmp_in), axis=0)
            tmp_gt = np.repeat(path2gt[p], tmp_in.shape[0], axis=0)
            ground_truth = np.concatenate((ground_truth, tmp_gt), axis=0)
            tmp_id = np.repeat(p, tmp_in.shape[0], axis=0)
            identifiers = np.concatenate((identifiers, tmp_id), axis=0)

    # 2) Load Tensorflow model to extract VGGish features
    with tf.Graph().as_default(), tf.Session() as sess:
        vggish_slim.define_vggish_slim(training=False)
        vggish_slim.load_vggish_slim_checkpoint(sess, 'vggish_model.ckpt')
        features_tensor = sess.graph.get_tensor_by_name(vggish_params.INPUT_TENSOR_NAME)
        embedding_tensor = sess.graph.get_tensor_by_name(vggish_params.OUTPUT_TENSOR_NAME)
        extracted_feat = sess.run([embedding_tensor], feed_dict={features_tensor: input_data})
        feature = np.squeeze(np.asarray(extracted_feat))

    return [feature, ground_truth, identifiers] 

def _guess_surf_file(fl):
    # MGH/MGZ files
    try: return np.asarray(fsmgh.load(fl).dataobj).flatten()
    except Exception: pass
    # FreeSurfer Curv files
    try: return fsio.read_morph_data(fl)
    except Exception: pass
    # Nifti files
    try: return np.squeeze(nib.load(fl).dataobj)
    except Exception: pass
    raise ValueError('Could not determine filetype for: %s' % fl) 

def gifti_to_array(gii):
    '''
    gifti_to_array(gii) yields the squeezed array of data contained in the given gifti object, gii,
      Note that if gii does not contain simple data in its darray object, then this will produce
      undefined results. This operation is effectively equivalent to:
      np.squeeze([x.data for x in gii.darrays]).
    gifti_to_array(gii_filename) is equivalent to gifti_to_array(neyropythy.load(gii_filename)).
    '''
    if pimms.is_str(gii): return gifti_to_array(ny.load(gii, 'gifti'))
    elif pimms.is_nparray(gii): return gii #already done
    elif isinstance(gii, nib.gifti.gifti.GiftiImage):
        return np.squeeze(np.asarray([x.data for x in gii.darrays]))
    else: raise ValueError('Could not understand argument to gifti_to_array') 

def load_cifti(filename, to='auto'):
    '''
    load_cifti(filename) yields the cifti image referened by the given filename by using the nibabel
      load function.
    
    The optional argument to may be used to coerce the resulting data to a particular format; the
    following arguments are understood:
      * 'header' will yield just the image header
      * 'data' will yield the image's data-array
      * 'field' will yield a squeezed version of the image's data-array and will raise an error if
        the data object has more than 2 non-unitary dimensions (appropriate for loading surface
        properties stored in image files)
      * 'image' will yield the raw image object
      * 'auto' is equivalent to 'image' unless the image has no more than 2 non-unitary dimensions,
        in which case it is assumed to be a surface-field and the return value is equivalent to
        the 'field' value.
    '''
    img = nib.load(filename)
    if not isinstance(img, nib.cifti2.Cifti2Image):
        raise ValueError('given file is not a cifti image')
    to = 'auto' if to is None else to.lower()
    if   to == 'image':  return img
    elif to == 'data':   return img.dataobj
    elif to == 'header': return img.header
    elif to == 'field':
        dat = np.squeeze(np.asarray(img.dataobj))
        if len(dat.shape) > 2:
            raise ValueError('image requested as field has more than 2 non-unitary dimensions')
        return dat
    elif to in ['auto', 'automatic']:
        dims = set(np.shape(img.dataobj))
        if 1 < len(dims) < 4 and 1 in dims: return np.squeeze(np.asarray(img.dataobj))
        else:                               return img
    else:
        raise ValueError('unrecognized \'to\' argument \'%s\'' % to) 

def load_mgh(filename, to='auto'):
    '''
    load_mgh(filename) yields the MGHImage referened by the given filename by using the
      nibabel.freesurfer.mghformat.load function.
    
    The optional argument 'to' may be used to coerce the resulting data to a particular format; the
    following arguments are understood:
      * 'header' will yield just the image header
      * 'data' will yield the image's data-array
      * 'field' will yield a squeezed version of the image's data-array and will raise an error if
        the data object has more than 2 non-unitary dimensions (appropriate for loading surface
        properties stored in image files)
      * 'affine' will yield the image's affine transformation
      * 'image' will yield the raw image object
      * 'auto' is equivalent to 'image' unless the image has no more than 2 non-unitary dimensions,
        in which case it is assumed to be a surface-field and the return value is equivalent to
        the 'field' value.
    '''
    img = fsmgh.load(filename)
    to = to.lower()
    if to == 'image':    return img
    elif to == 'data':   return img.dataobj
    elif to == 'affine': return img.affine
    elif to == 'header': return img.header
    elif to == 'field':
        dat = np.squeeze(img.dataobj)
        if len(dat.shape) > 2:
            raise ValueError('image requested as field has more than 2 non-unitary dimensions')
        return dat
    elif to in ['auto', 'automatic']:
        dims = set(img.dataobj.shape)
        if 1 < len(dims) < 4 and 1 in dims:
            return np.squeeze(img.dataobj)
        else:
            return img
    else:
        raise ValueError('unrecognized \'to\' argument \'%s\'' % to) 

def __call__(self, params):
        '''
        pf(params) yields the tuple (z, dz) where z is the potential value at the given parameters
          vector, params, and dz is the vector of the potential gradient.
        '''
        z  = self.value(params)
        dz = self.jacobian(params)
        if sps.issparse(dz): dz = dz.toarray()
        z  = np.squeeze(z)
        dz = np.squeeze(dz)
        return (z,dz) 

def jac(self):
        '''
        pf.jac() yields a jacobian calculation function for the given potential function pf that is
          appropritate for passing to a minimizer.
        '''
        def _jacobian(x):
            dz = self.jacobian(x)
            if sps.issparse(dz): dz = dz.toarray()
            dz = np.asarray(dz)
            return np.squeeze(dz)
        return _jacobian 

def load_nifti(filename, to='auto'):
    '''
    load_nifti(filename) yields the Nifti1Image or Nifti2Image referened by the given filename by
      using the nibabel load function.
    
    The optional argument to may be used to coerce the resulting data to a particular format; the
    following arguments are understood:
      * 'header' will yield just the image header
      * 'data' will yield the image's data-array
      * 'field' will yield a squeezed version of the image's data-array and will raise an error if
        the data object has more than 2 non-unitary dimensions (appropriate for loading surface
        properties stored in image files)
      * 'affine' will yield the image's affine transformation
      * 'image' will yield the raw image object
      * 'auto' is equivalent to 'image' unless the image has no more than 2 non-unitary dimensions,
        in which case it is assumed to be a surface-field and the return value is equivalent to
        the 'field' value.
    '''
    img = nib.load(filename)
    to = to.lower()
    if to == 'image':    return img
    elif to == 'data':   return img.dataobj
    elif to == 'affine': return img.affine
    elif to == 'header': return img.header
    elif to == 'field':
        dat = np.squeeze(np.asarray(img.dataobj))
        if len(dat.shape) > 2:
            raise ValueError('image requested as field has more than 2 non-unitary dimensions')
        return dat
    elif to in ['auto', 'automatic']:
        dims = set(np.shape(img.dataobj))
        if 1 < len(dims) < 4 and 1 in dims:
            return np.squeeze(np.asarray(img.dataobj))
        else:
            return img
    else:
        raise ValueError('unrecognized \'to\' argument \'%s\'' % to) 

def run_inference_on_image(image):
  """Runs inference on an image.

  Args:
    image: Image file name.

  Returns:
    Nothing
  """
  if not tf.gfile.Exists(image):
    tf.logging.fatal('File does not exist %s', image)
  image_data = tf.gfile.FastGFile(image, 'rb').read()

  # Creates graph from saved GraphDef.
  create_graph()

  with tf.Session() as sess:
    # Some useful tensors:
    # 'softmax:0': A tensor containing the normalized prediction across
    #   1000 labels.
    # 'pool_3:0': A tensor containing the next-to-last layer containing 2048
    #   float description of the image.
    # 'DecodeJpeg/contents:0': A tensor containing a string providing JPEG
    #   encoding of the image.
    # Runs the softmax tensor by feeding the image_data as input to the graph.
    softmax_tensor = sess.graph.get_tensor_by_name('softmax:0')
    predictions = sess.run(softmax_tensor,
                           {'DecodeJpeg/contents:0': image_data})
    predictions = np.squeeze(predictions)

    # Creates node ID --> English string lookup.
    node_lookup = NodeLookup()

    top_k = predictions.argsort()[-FLAGS.num_top_predictions:][::-1]
    for node_id in top_k:
      human_string = node_lookup.id_to_string(node_id)
      score = predictions[node_id]
      print('%s (score = %.5f)' % (human_string, score)) 

def test_diagonal_gradient_image(self):
    """Tests if a good pyramid image is created."""
    pyramid_image = test_utils.create_diagonal_gradient_image(3, 4, 2)

    # Test which is easy to understand.
    expected_first_channel = np.array([[3, 2, 1, 0],
                                       [4, 3, 2, 1],
                                       [5, 4, 3, 2]], dtype=np.float32)
    self.assertAllEqual(np.squeeze(pyramid_image[:, :, 0]),
                        expected_first_channel)

    # Actual test.
    expected_image = np.array([[[3, 30],
                                [2, 20],
                                [1, 10],
                                [0, 0]],
                               [[4, 40],
                                [3, 30],
                                [2, 20],
                                [1, 10]],
                               [[5, 50],
                                [4, 40],
                                [3, 30],
                                [2, 20]]], dtype=np.float32)

    self.assertAllEqual(pyramid_image, expected_image)