> -----Original Message-----
> From: ffmpeg-devel <ffmpeg-devel-boun...@ffmpeg.org> On Behalf Of Ting Fu
> Sent: Wednesday, July 29, 2020 10:11 PM
> To: ffmpeg-devel@ffmpeg.org
> Subject: [FFmpeg-devel] [PATCH V2 1/2] dnn/native: add native support for
> avg_pool
>
> Not support pooling strides in channel dimension now.
> It can be tested with the model generated with below python script:
>
> import tensorflow as tf
> import numpy as np
> import imageio
>
> in_img = imageio.imread('input_odd.jpg') in_img =
> in_img.astype(np.float32)/255.0 in_data = in_img[np.newaxis, :]
>
> x = tf.placeholder(tf.float32, shape=[1, None, None, 3], name='dnn_in') x_pool
> = tf.nn.avg_pool(x, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')
> #please
> alter the params as needed y = tf.identity(x_pool, name='dnn_out')
>
> sess=tf.Session()
> sess.run(tf.global_variables_initializer())
>
> graph_def = tf.graph_util.convert_variables_to_constants(sess, sess.graph_def,
> ['dnn_out']) tf.train.write_graph(graph_def, '.', 'image_process.pb',
> as_text=False)
>
> print("image_process.pb generated, please use \
> path_to_ffmpeg/tools/python/convert.py to generate image_process.model\n")
>
> output = sess.run(y, feed_dict={x: in_data}) imageio.imsave("out.jpg",
> np.squeeze(output))
>
> Signed-off-by: Ting Fu <ting...@intel.com>
> ---
> libavfilter/dnn/Makefile | 1 +
> libavfilter/dnn/dnn_backend_native.h | 2 +
> .../dnn/dnn_backend_native_layer_avgpool.c | 147 ++++++++++++++++++
> .../dnn/dnn_backend_native_layer_avgpool.h | 35 +++++
> .../dnn/dnn_backend_native_layer_conv2d.h | 3 +-
> libavfilter/dnn/dnn_backend_native_layers.c | 2 +
> tools/python/convert_from_tensorflow.py | 35 ++++-
> 7 files changed, 222 insertions(+), 3 deletions(-) create mode 100644
> libavfilter/dnn/dnn_backend_native_layer_avgpool.c
> create mode 100644 libavfilter/dnn/dnn_backend_native_layer_avgpool.h
>
> diff --git a/libavfilter/dnn/Makefile b/libavfilter/dnn/Makefile index
> d90137ec42..e0957073ee 100644
> --- a/libavfilter/dnn/Makefile
> +++ b/libavfilter/dnn/Makefile
> @@ -1,6 +1,7 @@
> OBJS-$(CONFIG_DNN) += dnn/dnn_interface.o
> OBJS-$(CONFIG_DNN) += dnn/dnn_backend_native.o
> OBJS-$(CONFIG_DNN) +=
> dnn/dnn_backend_native_layers.o
> +OBJS-$(CONFIG_DNN) +=
> dnn/dnn_backend_native_layer_avgpool.o
> OBJS-$(CONFIG_DNN) +=
> dnn/dnn_backend_native_layer_pad.o
> OBJS-$(CONFIG_DNN) +=
> dnn/dnn_backend_native_layer_conv2d.o
> OBJS-$(CONFIG_DNN) +=
> dnn/dnn_backend_native_layer_depth2space.o
[...]
>
>
> + def dump_avg_pool_to_file(self, node, f):
> + assert(node.op == 'AvgPool')
> + self.layer_number = self.layer_number + 1
> + self.converted_nodes.add(node.name)
> + node0 = self.name_node_dict[node.input[0]]
> + strides = node.attr['strides']
> + assert(strides.list.i[1]==strides.list.i[2])
> + assert(strides.list.i[0]==1)
> + assert(strides.list.i[3]==1)
Since the tensorflow do not support pooling strides in batch dimension, and
current do not support pooling in channel dimension,
added two assert here.
> + strides = strides.list.i[1]
> + filter_node = node.attr['ksize']
> + input_name = node.input[0]
> +
> + assert(filter_node.list.i[0]==1)
> + assert(filter_node.list.i[3]==1)
Same as above, the tensorflow do not support pooling ksize in both batch
dimension and channel dimension.
> + filter_height = filter_node.list.i[1]
> + filter_width = filter_node.list.i[2]
> +
> + in_channels = node0.attr['shape'].shape.dim[3].size
> + out_channels = in_channels
> + padding = node.attr['padding'].s.decode("utf-8")
> + np.array([self.op2code[node.op], strides,
> self.pool_paddings[padding],
> in_channels, out_channels,
> + filter_height],dtype=np.uint32).tofile(f)
> +
> + input_operand_index = self.add_operand(input_name,
> Operand.IOTYPE_INPUT)
> + output_operand_index = self.add_operand(node.name,
> Operand.IOTYPE_OUTPUT)
> + np.array([input_operand_index,
> + output_operand_index],dtype=np.uint32).tofile(f)
> +
> +
> def dump_layers_to_file(self, f):
> for node in self.nodes:
> if node.name in self.converted_nodes:
> @@ -311,6 +342,8 @@ class TFConverter:
>
> if node.op == 'Conv2D':
> self.dump_simple_conv2d_to_file(node, f)
> + if node.op == 'AvgPool':
> + self.dump_avg_pool_to_file(node, f)
> elif node.op == 'DepthToSpace':
> self.dump_depth2space_to_file(node, f)
> elif node.op == 'MirrorPad':
> --
> 2.17.1
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org
> with subject "unsubscribe".
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
