Python tensorflow.crop_and_resize() Examples

The following are 6 code examples of tensorflow.crop_and_resize(). 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 tensorflow , or try the search function .
Example #1
Source File: anchor_projector.py    From avod with MIT License 6 votes vote down vote up 
def reorder_projected_boxes(box_corners):
    """Helper function to reorder image corners.

    This reorders the corners from [x1, y1, x2, y2] to
    [y1, x1, y2, x2] which is required by the tf.crop_and_resize op.

    Args:
        box_corners: tensor image corners in the format
            N x [x1, y1, x2, y2]

    Returns:
        box_corners_reordered: tensor image corners in the format
            N x [y1, x1, y2, x2]
    """
    boxes_reordered = tf.stack([box_corners[:, 1],
                                box_corners[:, 0],
                                box_corners[:, 3],
                                box_corners[:, 2]],
                               axis=1)
    return boxes_reordered
Example #2
Source File: anchor_projector.py    From TLNet with Apache License 2.0 6 votes vote down vote up 
def reorder_projected_boxes(box_corners):
    """Helper function to reorder image corners.

    This reorders the corners from [x1, y1, x2, y2] to
    [y1, x1, y2, x2] which is required by the tf.crop_and_resize op.

    Args:
        box_corners: tensor image corners in the format
            N x [x1, y1, x2, y2]

    Returns:
        box_corners_reordered: tensor image corners in the format
            N x [y1, x1, y2, x2]
    """
    boxes_reordered = tf.stack([box_corners[:, 1],
                                box_corners[:, 0],
                                box_corners[:, 3],
                                box_corners[:, 2]],
                               axis=1)
    return boxes_reordered
Example #3
Source File: anchor_projector.py    From avod-ssd with MIT License 6 votes vote down vote up 
def reorder_projected_boxes(box_corners):
    """Helper function to reorder image corners.

    This reorders the corners from [x1, y1, x2, y2] to
    [y1, x1, y2, x2] which is required by the tf.crop_and_resize op.

    Args:
        box_corners: tensor image corners in the format
            N x [x1, y1, x2, y2]

    Returns:
        box_corners_reordered: tensor image corners in the format
            N x [y1, x1, y2, x2]
    """
    boxes_reordered = tf.stack([box_corners[:, 1],
                                box_corners[:, 0],
                                box_corners[:, 3],
                                box_corners[:, 2]],
                               axis=1)
    return boxes_reordered
Example #4
Source File: roi_pool.py    From Table-Detection-using-Deep-learning with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def _roi_crop(self, roi_proposals, conv_feature_map, im_shape):
        # Get normalized bounding boxes.
        bboxes = self._get_bboxes(roi_proposals, im_shape)
        # Generate fake batch ids
        bboxes_shape = tf.shape(bboxes)
        batch_ids = tf.zeros((bboxes_shape[0], ), dtype=tf.int32)
        # Apply crop and resize with extracting a crop double the desired size.
        crops = tf.image.crop_and_resize(
            conv_feature_map, bboxes, batch_ids,
            [self._pooled_width * 2, self._pooled_height * 2], name="crops"
        )

        # Applies max pool with [2,2] kernel to reduce the crops to half the
        # size, and thus having the desired output.
        prediction_dict = {
            'roi_pool': tf.nn.max_pool(
                crops, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1],
                padding=self._pooled_padding
            ),
        }

        if self._debug:
            prediction_dict['bboxes'] = bboxes
            prediction_dict['crops'] = crops
            prediction_dict['batch_ids'] = batch_ids
            prediction_dict['conv_feature_map'] = conv_feature_map

        return prediction_dict
Example #5
Source File: roi_pool.py    From Tabulo with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def _roi_crop(self, roi_proposals, conv_feature_map, im_shape):
        # Get normalized bounding boxes.
        bboxes = self._get_bboxes(roi_proposals, im_shape)
        # Generate fake batch ids
        bboxes_shape = tf.shape(bboxes)
        batch_ids = tf.zeros((bboxes_shape[0], ), dtype=tf.int32)
        # Apply crop and resize with extracting a crop double the desired size.
        crops = tf.image.crop_and_resize(
            conv_feature_map, bboxes, batch_ids,
            [self._pooled_width * 2, self._pooled_height * 2], name="crops"
        )

        # Applies max pool with [2,2] kernel to reduce the crops to half the
        # size, and thus having the desired output.
        prediction_dict = {
            'roi_pool': tf.nn.max_pool(
                crops, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1],
                padding=self._pooled_padding
            ),
        }

        if self._debug:
            prediction_dict['bboxes'] = bboxes
            prediction_dict['crops'] = crops
            prediction_dict['batch_ids'] = batch_ids
            prediction_dict['conv_feature_map'] = conv_feature_map

        return prediction_dict
Example #6
Source File: roi_pool.py    From luminoth with BSD 3-Clause "New" or "Revised" License 5 votes vote down vote up 
def _roi_crop(self, roi_proposals, conv_feature_map, im_shape):
        # Get normalized bounding boxes.
        bboxes = self._get_bboxes(roi_proposals, im_shape)
        # Generate fake batch ids
        bboxes_shape = tf.shape(bboxes)
        batch_ids = tf.zeros((bboxes_shape[0], ), dtype=tf.int32)
        # Apply crop and resize with extracting a crop double the desired size.
        crops = tf.image.crop_and_resize(
            conv_feature_map, bboxes, batch_ids,
            [self._pooled_width * 2, self._pooled_height * 2], name="crops"
        )

        # Applies max pool with [2,2] kernel to reduce the crops to half the
        # size, and thus having the desired output.
        prediction_dict = {
            'roi_pool': tf.nn.max_pool(
                crops, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1],
                padding=self._pooled_padding
            ),
        }

        if self._debug:
            prediction_dict['bboxes'] = bboxes
            prediction_dict['crops'] = crops
            prediction_dict['batch_ids'] = batch_ids
            prediction_dict['conv_feature_map'] = conv_feature_map

        return prediction_dict