Python fast_rcnn.config.cfg.PIXEL_MEANS Examples

The following are 30 code examples of fast_rcnn.config.cfg.PIXEL_MEANS(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module fast_rcnn.config.cfg , or try the search function .
Example #1
Source File: minibatch.py    From Faster-RCNN_TF with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
                                        cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #2
Source File: minibatch.py    From rgz_rcnn with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
                                        cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #3
Source File: torch_image_transform_layer.py    From py-R-FCN with MIT License 6 votes vote down vote up 
def setup(self, bottom, top):
        # (1, 3, 1, 1) shaped arrays
        self.PIXEL_MEANS = \
            np.array([[[[0.48462227599918]],
                       [[0.45624044862054]],
                       [[0.40588363755159]]]])
        self.PIXEL_STDS = \
            np.array([[[[0.22889466674951]],
                       [[0.22446679341259]],
                       [[0.22495548344775]]]])
        # The default ("old") pixel means that were already subtracted
        channel_swap = (0, 3, 1, 2)
        self.OLD_PIXEL_MEANS = \
            cfg.PIXEL_MEANS[np.newaxis, :, :, :].transpose(channel_swap)

        top[0].reshape(*(bottom[0].shape))
Example #4
Source File: torch_image_transform_layer.py    From face-py-faster-rcnn with MIT License 6 votes vote down vote up 
def setup(self, bottom, top):
        # (1, 3, 1, 1) shaped arrays
        self.PIXEL_MEANS = \
            np.array([[[[0.48462227599918]],
                       [[0.45624044862054]],
                       [[0.40588363755159]]]])
        self.PIXEL_STDS = \
            np.array([[[[0.22889466674951]],
                       [[0.22446679341259]],
                       [[0.22495548344775]]]])
        # The default ("old") pixel means that were already subtracted
        channel_swap = (0, 3, 1, 2)
        self.OLD_PIXEL_MEANS = \
            cfg.PIXEL_MEANS[np.newaxis, :, :, :].transpose(channel_swap)

        top[0].reshape(*(bottom[0].shape))
Example #5
Source File: minibatch.py    From scene-graph-TF-release with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = roidb[i]['image']() # use image getter

        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
                                        cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #6
Source File: minibatch.py    From rgz_rcnn with MIT License 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, overlaps):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[1:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' overlap: ', overlaps[i]
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #7
Source File: minibatch.py    From SubCNN with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]

        im_orig = im.astype(np.float32, copy=True)
        im_orig -= cfg.PIXEL_MEANS

        im_scale = cfg.TRAIN.SCALES_BASE[scale_inds[i]]
        im = cv2.resize(im_orig, None, None, fx=im_scale, fy=im_scale, interpolation=cv2.INTER_LINEAR)

        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #8
Source File: minibatch.py    From faster-rcnn-resnet with MIT License 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, overlaps):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[1:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' overlap: ', overlaps[i]
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #9
Source File: minibatch.py    From faster-rcnn-resnet with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
                                        cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #10
Source File: minibatch.py    From py-R-FCN with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
                                        cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #11
Source File: torch_image_transform_layer.py    From faster-rcnn-resnet with MIT License 6 votes vote down vote up 
def setup(self, bottom, top):
        # (1, 3, 1, 1) shaped arrays
        self.PIXEL_MEANS = \
            np.array([[[[0.48462227599918]],
                       [[0.45624044862054]],
                       [[0.40588363755159]]]])
        self.PIXEL_STDS = \
            np.array([[[[0.22889466674951]],
                       [[0.22446679341259]],
                       [[0.22495548344775]]]])
        # The default ("old") pixel means that were already subtracted
        channel_swap = (0, 3, 1, 2)
        self.OLD_PIXEL_MEANS = \
            cfg.PIXEL_MEANS[np.newaxis, :, :, :].transpose(channel_swap)

        top[0].reshape(*(bottom[0].shape))
Example #12
Source File: minibatch.py    From py-R-FCN with MIT License 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, overlaps):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[1:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' overlap: ', overlaps[i]
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #13
Source File: minibatch.py    From face-py-faster-rcnn with MIT License 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, overlaps):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[1:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' overlap: ', overlaps[i]
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #14
Source File: minibatch.py    From face-py-faster-rcnn with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
                                        cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #15
Source File: minibatch.py    From SubCNN with MIT License 6 votes vote down vote up 
def _process_images(roidb):
    """Builds an input blob from the images in the roidb
    """
    num_images = len(roidb)
    processed_ims = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]

        im_orig = im.astype(np.float32, copy=True)
        im_orig -= cfg.PIXEL_MEANS

        processed_ims.append(im_orig)

    return processed_ims
Example #16
Source File: minibatch.py    From oicr with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    im_shapes = np.zeros((0, 2), dtype=np.float32)
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale, im_shape = prep_im_for_blob(im, cfg.PIXEL_MEANS, 
                                                  target_size, 
                                                  cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)
        im_shapes = np.vstack((im_shapes, im_shape))

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales, im_shapes
Example #17
Source File: minibatch.py    From Faster-RCNN_TF with MIT License 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, overlaps):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[1:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' overlap: ', overlaps[i]
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #18
Source File: minibatch2.py    From Faster-RCNN_TF with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]

        im_orig = im.astype(np.float32, copy=True)
        im_orig -= cfg.PIXEL_MEANS

        im_scale = cfg.TRAIN.SCALES_BASE[scale_inds[i]]
        im = cv2.resize(im_orig, None, None, fx=im_scale, fy=im_scale, interpolation=cv2.INTER_LINEAR)

        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #19
Source File: minibatch2.py    From Faster-RCNN_TF with MIT License 6 votes vote down vote up 
def _get_image_blob_multiscale(roidb):
    """Builds an input blob from the images in the roidb at multiscales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    scales = cfg.TRAIN.SCALES_BASE
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]

        im_orig = im.astype(np.float32, copy=True)
        im_orig -= cfg.PIXEL_MEANS

        for im_scale in scales:
            im = cv2.resize(im_orig, None, None, fx=im_scale, fy=im_scale, interpolation=cv2.INTER_LINEAR)
            im_scales.append(im_scale)
            processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #20
Source File: minibatch.py    From face-magnet with Apache License 2.0 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, overlaps):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[1:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' overlap: ', overlaps[i]
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
        )
        plt.show()
Example #21
Source File: minibatch.py    From Faster-RCNN_TF with MIT License 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, sublabels_blob):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[2:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        subcls = sublabels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' subclass: ', subcls
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #22
Source File: minibatch.py    From caffe-faster-rcnn-resnet-fpn with MIT License 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, overlaps):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[1:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' overlap: ', overlaps[i]
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #23
Source File: torch_image_transform_layer.py    From uai-sdk with Apache License 2.0 6 votes vote down vote up 
def setup(self, bottom, top):
        # (1, 3, 1, 1) shaped arrays
        self.PIXEL_MEANS = \
            np.array([[[[0.48462227599918]],
                       [[0.45624044862054]],
                       [[0.40588363755159]]]])
        self.PIXEL_STDS = \
            np.array([[[[0.22889466674951]],
                       [[0.22446679341259]],
                       [[0.22495548344775]]]])
        # The default ("old") pixel means that were already subtracted
        channel_swap = (0, 3, 1, 2)
        self.OLD_PIXEL_MEANS = \
            cfg.PIXEL_MEANS[np.newaxis, :, :, :].transpose(channel_swap)

        top[0].reshape(*(bottom[0].shape))
Example #24
Source File: minibatch.py    From caffe-faster-rcnn-resnet-fpn with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
                                        cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #25
Source File: minibatch.py    From uai-sdk with Apache License 2.0 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
                                        cfg.TRAIN.MAX_SIZE)
        im_scales.append(im_scale)
        processed_ims.append(im)

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales
Example #26
Source File: minibatch.py    From uai-sdk with Apache License 2.0 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, overlaps):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[1:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' overlap: ', overlaps[i]
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #27
Source File: minibatch.py    From SubCNN with MIT License 6 votes vote down vote up 
def _vis_minibatch(im_blob, rois_blob, labels_blob, sublabels_blob):
    """Visualize a mini-batch for debugging."""
    import matplotlib.pyplot as plt
    for i in xrange(rois_blob.shape[0]):
        rois = rois_blob[i, :]
        im_ind = rois[0]
        roi = rois[2:]
        im = im_blob[im_ind, :, :, :].transpose((1, 2, 0)).copy()
        im += cfg.PIXEL_MEANS
        im = im[:, :, (2, 1, 0)]
        im = im.astype(np.uint8)
        cls = labels_blob[i]
        subcls = sublabels_blob[i]
        plt.imshow(im)
        print 'class: ', cls, ' subclass: ', subcls
        plt.gca().add_patch(
            plt.Rectangle((roi[0], roi[1]), roi[2] - roi[0],
                          roi[3] - roi[1], fill=False,
                          edgecolor='r', linewidth=3)
            )
        plt.show()
Example #28
Source File: minibatch.py    From dpl with MIT License 6 votes vote down vote up 
def _get_image_blob(roidb, scale_inds):
    """Builds an input blob from the images in the roidb at the specified
    scales.
    """
    num_images = len(roidb)
    processed_ims = []
    im_scales = []
    im_shapes = np.zeros((0, 2), dtype=np.float32)
    for i in xrange(num_images):
        im = cv2.imread(roidb[i]['image'])
        if roidb[i]['flipped']:
            im = im[:, ::-1, :]
        target_size = cfg.TRAIN.SCALES[scale_inds[i]]
        im, im_scale, im_shape = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size)
        im_scales.append(im_scale)
        processed_ims.append(im)
        im_shapes = np.vstack((im_shapes, im_shape))

    # Create a blob to hold the input images
    blob = im_list_to_blob(processed_ims)

    return blob, im_scales, im_shapes
Example #29
Source File: torch_image_transform_layer.py    From face-magnet with Apache License 2.0 6 votes vote down vote up 
def setup(self, bottom, top):
        # (1, 3, 1, 1) shaped arrays
        self.PIXEL_MEANS = \
            np.array([[[[0.48462227599918]],
                       [[0.45624044862054]],
                       [[0.40588363755159]]]])
        self.PIXEL_STDS = \
            np.array([[[[0.22889466674951]],
                       [[0.22446679341259]],
                       [[0.22495548344775]]]])
        # The default ("old") pixel means that were already subtracted
        channel_swap = (0, 3, 1, 2)
        self.OLD_PIXEL_MEANS = \
            cfg.PIXEL_MEANS[np.newaxis, :, :, :].transpose(channel_swap)

        top[0].reshape(*(bottom[0].shape))
Example #30
Source File: torch_image_transform_layer.py    From caffe-faster-rcnn-resnet-fpn with MIT License 6 votes vote down vote up 
def setup(self, bottom, top):
        # (1, 3, 1, 1) shaped arrays
        self.PIXEL_MEANS = \
            np.array([[[[0.48462227599918]],
                       [[0.45624044862054]],
                       [[0.40588363755159]]]])
        self.PIXEL_STDS = \
            np.array([[[[0.22889466674951]],
                       [[0.22446679341259]],
                       [[0.22495548344775]]]])
        # The default ("old") pixel means that were already subtracted
        channel_swap = (0, 3, 1, 2)
        self.OLD_PIXEL_MEANS = \
            cfg.PIXEL_MEANS[np.newaxis, :, :, :].transpose(channel_swap)

        top[0].reshape(*(bottom[0].shape))