more math unary operations will be added here It can be tested with the model file generated with below python scripy: import tensorflow as tf import numpy as np import imageio in_img = imageio.imread('input.jpeg') 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.subtract(x, 0.5) x2 = tf.abs(x1) 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))
Showing the changes to tools/python/convert_from_tensorflow.py (1 of 7 files). Show all
Unified · Split
| @@ -70,8 +70,9 @@ class TFConverter: · 69 unmodified lines | ||
| 70 | 70 | self.converted_nodes = set() |
| 71 | 71 | self.conv2d_scope_names = set() |
| 72 | 72 | self.conv2d_scopename_inputname_dict = {} |
| 73 | − self.op2code = {'Conv2D':1, 'DepthToSpace':2, 'MirrorPad':3, 'Maximum':4, 'MathBinary':5} | |
| 73 | + self.op2code = {'Conv2D':1, 'DepthToSpace':2, 'MirrorPad':3, 'Maximum':4, 'MathBinary':5, 'MathUnary':6} | |
| 74 | 74 | self.mathbin2code = {'Sub':0, 'Add':1, 'Mul':2, 'RealDiv':3, 'Minimum':4} |
| 75 | + self.mathun2code = {'Abs':0} | |
| 75 | 76 | self.mirrorpad_mode = {'CONSTANT':0, 'REFLECT':1, 'SYMMETRIC':2} |
| 76 | 77 | self.name_operand_dict = {} |
| 77 | 78 | |
| @@ -286,6 +287,17 @@ class TFConverter: · 208 unmodified lines | ||
| 286 | 287 | np.array([output_operand_index], dtype=np.uint32).tofile(f) |
| 287 | 288 | |
| 288 | 289 | |
| 290 | + def dump_mathunary_to_file(self, node, f): | |
| 291 | + self.layer_number = self.layer_number + 1 | |
| 292 | + self.converted_nodes.add(node.name) | |
| 293 | + i0_node = self.name_node_dict[node.input[0]] | |
| 294 | + np.array([self.op2code['MathUnary'], self.mathun2code[node.op]], dtype=np.uint32).tofile(f) | |
| 295 | + input_operand_index = self.add_operand(i0_node.name, Operand.IOTYPE_INPUT) | |
| 296 | + np.array([input_operand_index], dtype=np.uint32).tofile(f) | |
| 297 | + output_operand_index = self.add_operand(node.name, Operand.IOTYPE_OUTPUT) | |
| 298 | + np.array([output_operand_index],dtype=np.uint32).tofile(f) | |
| 299 | + | |
| 300 | + | |
| 289 | 301 | def dump_layers_to_file(self, f): |
| 290 | 302 | for node in self.nodes: |
| 291 | 303 | if node.name in self.converted_nodes: |
| @@ -307,6 +319,8 @@ class TFConverter: · 15 unmodified lines | ||
| 307 | 319 | self.dump_maximum_to_file(node, f) |
| 308 | 320 | elif node.op in self.mathbin2code: |
| 309 | 321 | self.dump_mathbinary_to_file(node, f) |
| 322 | + elif node.op in self.mathun2code: | |
| 323 | + self.dump_mathunary_to_file(node, f) | |
| 310 | 324 | |
| 311 | 325 | |
| 312 | 326 | def dump_operands_to_file(self, f): |