dnn_backend_native_layer_mathunary: add floor support

It can be tested with the model generated with below python script:

import tensorflow as tf
import os
import numpy as np
import imageio
from tensorflow.python.framework import graph_util
name = 'floor'

pb_file_path = os.getcwd()
if not os.path.exists(pb_file_path+'/{}_savemodel/'.format(name)):
    os.mkdir(pb_file_path+'/{}_savemodel/'.format(name))

with tf.Session(graph=tf.Graph()) as sess:
    in_img = imageio.imread('detection.jpg')
    in_img = in_img.astype(np.float32)
    in_data = in_img[np.newaxis, :]
    input_x = tf.placeholder(tf.float32, shape=[1, None, None, 3], name='dnn_in')
    y_ = tf.math.floor(input_x*255)/255
    y = tf.identity(y_, name='dnn_out')
    sess.run(tf.global_variables_initializer())
    constant_graph = graph_util.convert_variables_to_constants(sess, sess.graph_def, ['dnn_out'])

    with tf.gfile.FastGFile(pb_file_path+'/{}_savemodel/model.pb'.format(name), mode='wb') as f:
        f.write(constant_graph.SerializeToString())

    print("model.pb generated, please in ffmpeg path use\n \n \
    python tools/python/convert.py {}_savemodel/model.pb --outdir={}_savemodel/ \n \nto generate model.model\n".format(name,name))

    output = sess.run(y, feed_dict={ input_x: in_data})
    imageio.imsave("out.jpg", np.squeeze(output))

    print("To verify, please ffmpeg path use\n \n \
    ./ffmpeg -i detection.jpg -vf format=rgb24,dnn_processing=model={}_savemodel/model.pb:input=dnn_in:output=dnn_out:dnn_backend=tensorflow -f framemd5 {}_savemodel/tensorflow_out.md5\n  \
    or\n \
    ./ffmpeg -i detection.jpg -vf format=rgb24,dnn_processing=model={}_savemodel/model.pb:input=dnn_in:output=dnn_out:dnn_backend=tensorflow {}_savemodel/out_tensorflow.jpg\n \nto generate output result of tensorflow model\n".format(name, name, name, name))

    print("To verify, please ffmpeg path use\n \n \
    ./ffmpeg -i detection.jpg -vf format=rgb24,dnn_processing=model={}_savemodel/model.model:input=dnn_in:output=dnn_out:dnn_backend=native -f framemd5 {}_savemodel/native_out.md5\n  \
    or \n \
    ./ffmpeg -i detection.jpg -vf format=rgb24,dnn_processing=model={}_savemodel/model.model:input=dnn_in:output=dnn_out:dnn_backend=native {}_savemodel/out_native.jpg\n \nto generate output result of native model\n".format(name, name, name, name))

Signed-off-by: Mingyu Yin <mingyu.yin@intel.com>

diff --git a/libavfilter/dnn/dnn_backend_native_layer_mathunary.c b/libavfilter/dnn/dnn_backend_native_layer_mathunary.c
index a62f6ba6f05ecf71b7105941d04652e3ec76fba7..f9f88983f9062efbf6597093be424f087aad6430 100644 (file)
--- a/libavfilter/dnn/dnn_backend_native_layer_mathunary.c
+++ b/libavfilter/dnn/dnn_backend_native_layer_mathunary.c
@@ -134,6 +134,10 @@ int dnn_execute_layer_math_unary(DnnOperand *operands, const int32_t *input_oper
         for (int i = 0; i < dims_count; ++i)
             dst[i] = ceil(src[i]);
         return 0;
+    case DMUO_FLOOR:
+        for (int i = 0; i < dims_count; ++i)
+            dst[i] = floor(src[i]);
+        return 0;
     default:
         return -1;
     }

diff --git a/libavfilter/dnn/dnn_backend_native_layer_mathunary.h b/libavfilter/dnn/dnn_backend_native_layer_mathunary.h
index 82b2d7f4ab912b58256c0344425bc89ad55d8f0b..1b78fbe1c214d153935fe9968b82defae2613c2c 100644 (file)
--- a/libavfilter/dnn/dnn_backend_native_layer_mathunary.h
+++ b/libavfilter/dnn/dnn_backend_native_layer_mathunary.h
@@ -44,6 +44,7 @@ typedef enum {
     DMUO_ACOSH = 11,
     DMUO_ATANH = 12,
     DMUO_CEIL = 13,
+    DMUO_FLOOR = 14,
     DMUO_COUNT
 } DNNMathUnaryOperation;
 

diff --git a/tests/dnn/dnn-layer-mathunary-test.c b/tests/dnn/dnn-layer-mathunary-test.c
index 7da3a206ed3a2d3a7946dbe3386763f9a74f36f1..0bfcdcbd8585dc120a5a30f7cca860a9a6d4a7c5 100644 (file)
--- a/tests/dnn/dnn-layer-mathunary-test.c
+++ b/tests/dnn/dnn-layer-mathunary-test.c
@@ -58,6 +58,8 @@ static float get_expected(float f, DNNMathUnaryOperation op)
         return atanh(f);
     case DMUO_CEIL:
         return ceil(f);
+    case DMUO_FLOOR:
+        return floor(f);
     default:
         av_assert0(!"not supported yet");
         return 0.f;
@@ -132,5 +134,7 @@ int main(int agrc, char **argv)
         return 1;
     if (test(DMUO_CEIL))
         return 1;
+    if (test(DMUO_FLOOR))
+        return 1;
     return 0;
 }

diff --git a/tools/python/convert_from_tensorflow.py b/tools/python/convert_from_tensorflow.py
index 64b75513148d830612bce98da398dd9dcdd0e705..2fb8c4c303b7d88dc2b232ba9fbce2935d5042fe 100644 (file)
--- a/tools/python/convert_from_tensorflow.py
+++ b/tools/python/convert_from_tensorflow.py
@@ -74,7 +74,7 @@ class TFConverter:
         self.mathbin2code = {'Sub':0, 'Add':1, 'Mul':2, 'RealDiv':3, 'Minimum':4}
         self.mathun2code  = {'Abs':0, 'Sin':1, 'Cos':2, 'Tan':3, 'Asin':4,
                 'Acos':5, 'Atan':6, 'Sinh':7, 'Cosh':8, 'Tanh':9, 'Asinh':10,
-                'Acosh':11, 'Atanh':12, 'Ceil':13}
+                'Acosh':11, 'Atanh':12, 'Ceil':13, 'Floor':14}
         self.mirrorpad_mode = {'CONSTANT':0, 'REFLECT':1, 'SYMMETRIC':2}
         self.name_operand_dict = {}
 

diff --git a/tools/python/convert_header.py b/tools/python/convert_header.py
index 62f1d342f35d87d241be615d89bb9370d3fba22a..b12b5079fc8ea9b84e6c62ce991145b84ae16e86 100644 (file)
--- a/tools/python/convert_header.py
+++ b/tools/python/convert_header.py
@@ -23,4 +23,4 @@ str = 'FFMPEGDNNNATIVE'
 major = 1
 
 # increase minor when we don't have to re-convert the model file
-minor = 19
+minor = 20