U^2-Net bugfixed, tflite,coreml,tfjs,tftrt

061_U-2-Net/01_float32/04_int_fullint_quant.py

@@ -0,0 +1,46 @@
+### tensorflow==2.3.1
+
+import tensorflow as tf
+import numpy as np
+import sys
+import tensorflow_datasets as tfds
+
+def representative_dataset_gen():
+    for data in raw_test_data.take(10):
+        image = data['image'].numpy()
+        image = tf.image.resize(image, (height, width))
+        image = image[np.newaxis,:,:,:]
+        image = image / 127.5 - 1.0
+        yield [image]
+
+raw_test_data, info = tfds.load(name="coco/2017", with_info=True, split="test", data_dir="~/TFDS", download=False)
+
+height = 480
+width  = 640
+
+# Integer Quantization - Input/Output=float32
+converter = tf.lite.TFLiteConverter.from_saved_model('saved_model_{}x{}'.format(height, width))
+converter.optimizations = [tf.lite.Optimize.DEFAULT]
+converter.representative_dataset = representative_dataset_gen
+tflite_quant_model = converter.convert()
+with open('u2netp_{}x{}_integer_quant.tflite'.format(height, width), 'wb') as w:
+    w.write(tflite_quant_model)
+print('Integer Quantization complete! - u2netp_{}x{}_integer_quant.tflite'.format(height, width))
+
+
+# # Full Integer Quantization - Input/Output=int8
+# converter = tf.lite.TFLiteConverter.from_saved_model('saved_model_{}x{}'.format(height, width))
+# converter.optimizations = [tf.lite.Optimize.DEFAULT]
+# converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]
+# converter.inference_input_type = tf.uint8
+# converter.inference_output_type = tf.uint8
+# converter.representative_dataset = representative_dataset_gen
+# tflite_quant_model = converter.convert()
+# with open('u2netp_{}x{}_full_integer_quant.tflite'.format(height, width), 'wb') as w:
+#     w.write(tflite_quant_model)
+# print('Integer Quantization complete! - u2netp_{}x{}_full_integer_quant.tflite'.format(height, width))
+
+# # EdgeTPU
+# import subprocess
+# result = subprocess.check_output(["edgetpu_compiler", "-s", "u2netp_{}x{}_full_integer_quant.tflite".format(height, width)])
+# print(result)

061_U-2-Net/01_float32/04_u2net_bin2h5_weight_float16_quant.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+pip3 install openvino2tensorflow --upgrade
+
+openvino2tensorflow \
+--model_path openvino/256x256/FP32/u2netp_256x256.xml \
+--model_output_path saved_model_256x256 \
+--output_saved_model True \
+--output_h5 True \
+--output_pb True \
+--output_no_quant_float32_tflite True \
+--output_weight_quant_tflite True \
+--output_float16_quant_tflite True
+
+
+openvino2tensorflow \
+--model_path openvino/320x320/FP32/u2netp_320x320.xml \
+--model_output_path saved_model_320x320 \
+--output_saved_model True \
+--output_h5 True \
+--output_pb True \
+--output_no_quant_float32_tflite True \
+--output_weight_quant_tflite True \
+--output_float16_quant_tflite True
+
+
+openvino2tensorflow \
+--model_path openvino/480x640/FP32/u2netp_480x640.xml \
+--model_output_path saved_model_480x640 \
+--output_saved_model True \
+--output_h5 True \
+--output_pb True \
+--output_no_quant_float32_tflite True \
+--output_weight_quant_tflite True \
+--output_float16_quant_tflite True