it can be tested with model file generated with below python script: import tensorflow as tf import numpy as np import imageio in_img = imageio.imread('input.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') x1 = tf.minimum(0.7, x) x2 = tf.maximum(x1, 0.4) y = tf.identity(x2, 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))
Unified · Split
| @@ -150,6 +150,19 @@ int dnn_execute_layer_math_binary(DnnOperand *operands, const int32_t *input_ope · 149 unmodified lines | ||
| 150 | 150 | } |
| 151 | 151 | } |
| 152 | 152 | return 0; |
| 153 | + case DMBO_MINIMUM: | |
| 154 | + if (params->input0_broadcast || params->input1_broadcast) { | |
| 155 | + for (int i = 0; i < dims_count; ++i) { | |
| 156 | + dst[i] = FFMIN(params->v, src[i]); | |
| 157 | + } | |
| 158 | + } else { | |
| 159 | + const DnnOperand *input1 = &operands[input_operand_indexes[1]]; | |
| 160 | + const float *src1 = input1->data; | |
| 161 | + for (int i = 0; i < dims_count; ++i) { | |
| 162 | + dst[i] = FFMIN(src[i], src1[i]); | |
| 163 | + } | |
| 164 | + } | |
| 165 | + return 0; | |
| 153 | 166 | default: |
| 154 | 167 | return -1; |
| 155 | 168 | } |
| @@ -35,6 +35,7 @@ typedef enum { · 34 unmodified lines | ||
| 35 | 35 | DMBO_ADD = 1, |
| 36 | 36 | DMBO_MUL = 2, |
| 37 | 37 | DMBO_REALDIV = 3, |
| 38 | + DMBO_MINIMUM = 4, | |
| 38 | 39 | DMBO_COUNT |
| 39 | 40 | } DNNMathBinaryOperation; |
| 40 | 41 | |
| @@ -71,7 +71,7 @@ class TFConverter: · 70 unmodified lines | ||
| 71 | 71 | self.conv2d_scope_names = set() |
| 72 | 72 | self.conv2d_scopename_inputname_dict = {} |
| 73 | 73 | self.op2code = {'Conv2D':1, 'DepthToSpace':2, 'MirrorPad':3, 'Maximum':4, 'MathBinary':5} |
| 74 | − self.mathbin2code = {'Sub':0, 'Add':1, 'Mul':2, 'RealDiv':3} | |
| 74 | + self.mathbin2code = {'Sub':0, 'Add':1, 'Mul':2, 'RealDiv':3, 'Minimum':4} | |
| 75 | 75 | self.mirrorpad_mode = {'CONSTANT':0, 'REFLECT':1, 'SYMMETRIC':2} |
| 76 | 76 | self.name_operand_dict = {} |
| 77 | 77 | |
| @@ -305,15 +305,10 @@ class TFConverter: · 227 unmodified lines | ||
| 305 | 305 | self.dump_mirrorpad_to_file(node, f) |
| 306 | 306 | elif node.op == 'Maximum': |
| 307 | 307 | self.dump_maximum_to_file(node, f) |
| 308 | − elif node.op == 'Sub': | |
| 309 | − self.dump_mathbinary_to_file(node, f) | |
| 310 | − elif node.op == 'Add': | |
| 311 | − self.dump_mathbinary_to_file(node, f) | |
| 312 | − elif node.op == 'Mul': | |
| 313 | − self.dump_mathbinary_to_file(node, f) | |
| 314 | − elif node.op == 'RealDiv': | |
| 308 | + elif node.op in self.mathbin2code: | |
| 315 | 309 | self.dump_mathbinary_to_file(node, f) |
| 316 | 310 | |
| 311 | + | |
| 317 | 312 | def dump_operands_to_file(self, f): |
| 318 | 313 | operands = sorted(self.name_operand_dict.values()) |
| 319 | 314 | for operand in operands: |
| @@ -23,4 +23,4 @@ str = 'FFMPEGDNNNATIVE' · 22 unmodified lines | ||
| 23 | 23 | major = 1 |
| 24 | 24 | |
| 25 | 25 | # increase minor when we don't have to re-convert the model file |
| 26 | −minor = 4 | |
| 26 | +minor = 5 | |