#!/usr/bin/env python
# ******************************************************************************
# Copyright 2020 Brainchip Holdings Ltd.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ******************************************************************************
"""
YOLO model definition for detection.
"""
__all__ = ["yolo_base", "yolo_widerface_pretrained", "yolo_voc_pretrained"]
import pickle
import numpy as np
from keras import Model
from ..layer_blocks import yolo_head_block
from ..imagenet.model_akidanet import akidanet_imagenet
from ..utils import fetch_file
from ..model_io import load_model, get_model_path
[docs]def yolo_base(input_shape=(224, 224, 3), classes=1, nb_box=5, alpha=1.0, input_scaling=(127.5, -1)):
""" Instantiates the YOLOv2 architecture.
Args:
input_shape (tuple, optional): input shape tuple. Defaults to (224, 224, 3).
classes (int, optional): number of classes to classify images into. Defaults to 1.
nb_box (int, optional): number of anchors boxes to use. Defaults to 5.
alpha (float, optional): controls the width of the model. Defaults to 1.0.
input_scaling (tuple, optional): scale factor and offset to apply to
inputs. Defaults to (127.5, -1). Note that following Akida
convention, the scale factor is a number used as a divisor.
Returns:
keras.Model: a Keras Model instance.
"""
# Create an AkidaNet network without top layers
base_model = akidanet_imagenet(input_shape=input_shape, alpha=alpha, include_top=False,
input_scaling=input_scaling)
# Add YOLO top layers to the base model
input_shape = base_model.input_shape
x = yolo_head_block(base_model.layers[-1].output, num_boxes=nb_box, classes=classes)
model = Model(inputs=base_model.input, outputs=x, name='yolo_base')
# Initialize detection layer weights
layers = [layer for layer in model.layers if "detection_layer" in layer.name]
assert len(layers) in (1, 2), "No detection layer found."
if len(layers) == 1:
# sepconv is fused on Akida v1
layer = layers[0]
detection_weights = layer.get_weights()
dw_weights_shape = detection_weights[0].shape
pw_weights_shape = detection_weights[1].shape
bias_shape = detection_weights[2].shape
else:
# sepconv is unfused on Akida v2
dw_layer = layers[0]
pw_layer = layers[1]
dw_weights = dw_layer.get_weights()
assert len(dw_weights) == 1 # no bias
pw_weights = pw_layer.get_weights()
dw_weights_shape = dw_weights[0].shape
pw_weights_shape = pw_weights[0].shape
bias_shape = pw_weights[1].shape
mu, sigma = 0, 0.1
grid_size = model.output_shape[1:3]
grid_area = grid_size[0] * grid_size[1]
dw_kernel = np.random.normal(mu, sigma,
size=dw_weights_shape) / grid_area
pw_kernel = np.random.normal(mu, sigma,
size=pw_weights_shape) / grid_area
bias = np.random.normal(mu, sigma,
size=bias_shape) / grid_area
if len(layers) == 1:
layer.set_weights([dw_kernel, pw_kernel, bias])
else:
dw_layer.set_weights([dw_kernel])
pw_layer.set_weights([pw_kernel, bias])
return model
[docs]def yolo_widerface_pretrained(quantized=True):
"""
Helper method to retrieve a `yolo_base` model that was trained on WiderFace
dataset and the anchors that are needed to interpet the model output.
Args:
quantized (bool, optional): a boolean indicating whether the model should be loaded
quantized or not. Defaults to True.
Returns:
keras.Model, list: a Keras Model instance and a list of anchors.
"""
anchors_name = 'widerface_anchors.pkl'
anchors_path = fetch_file(
'https://data.brainchip.com/dataset-mirror/widerface/' + anchors_name,
fname=anchors_name,
file_hash='325f92336a310d83fed71765436ee343bf3e39cbc12fd099d30677761aee9376',
cache_subdir='datasets/widerface')
with open(anchors_path, 'rb') as handle:
anchors = pickle.load(handle)
if quantized:
model_name_v1 = 'yolo_akidanet_widerface_iq8_wq4_aq4.h5'
file_hash_v1 = 'd55744cbfbbe1131aa26015f38d99f1df7026347bae8f66683259e366e7b6e03'
model_name_v2 = 'yolo_akidanet_widerface_i8_w4_a4.h5'
file_hash_v2 = '8aa1277d705366e47908cfc9919bd1e92770185faee2ca6b33c37ac645663663'
else:
model_name_v1 = 'yolo_akidanet_widerface.h5'
file_hash_v1 = 'fc1eedb343a97b637f7877a9978d0230962360643578f5fe357034d653b37e44'
model_name_v2 = 'yolo_akidanet_widerface.h5'
file_hash_v2 = 'fdb28048d93548956616fe368b6b7f924650a558ea2ed9c3572e774948f6b6e1'
model_path, model_name, file_hash = get_model_path("yolo", model_name_v1, file_hash_v1,
model_name_v2, file_hash_v2)
model_path = fetch_file(model_path,
fname=model_name,
file_hash=file_hash,
cache_subdir='models')
return load_model(model_path), anchors
[docs]def yolo_voc_pretrained(quantized=True):
"""
Helper method to retrieve a `yolo_base` model that was trained on PASCAL
VOC2012 dataset for 'person' and 'car' classes only, and the anchors that
are needed to interpet the model output.
Args:
quantized (bool, optional): a boolean indicating whether the model should be loaded
quantized or not. Defaults to True.
Returns:
keras.Model, list: a Keras Model instance and a list of anchors.
"""
anchors_name = 'coco_anchors.pkl'
anchors_path = fetch_file(
'https://data.brainchip.com/dataset-mirror/coco/' + anchors_name,
fname=anchors_name,
file_hash='36993699182495dd843158583515bd8d1412da978c55286ba0fefa88f5a8cace',
cache_subdir='datasets/coco')
with open(anchors_path, 'rb') as handle:
anchors = pickle.load(handle)
if quantized:
model_name_v1 = 'yolo_akidanet_voc_iq8_wq4_aq4.h5'
file_hash_v1 = 'e65b0b6bd4b08c2796c3bbea89343748195e29e240ce28e70489c53d06ca69d9'
model_name_v2 = 'yolo_akidanet_voc_i8_w4_a4.h5'
file_hash_v2 = '60e6fbfac6dabe8509df5109d72462e21b53eae53741c619bbe89104fe4c7c4f'
else:
model_name_v1 = 'yolo_akidanet_voc.h5'
file_hash_v1 = '5dff1dd3afafd512e105fa416b444431ca5f816ccc42d9efb49cfa34bd13e91d'
model_name_v2 = 'yolo_akidanet_voc.h5'
file_hash_v2 = 'e680e43a136ed3ba23580f4f4bf02be01fdf53d2675fec9c51116e9e79c083ae'
model_path, model_name, file_hash = get_model_path("yolo", model_name_v1, file_hash_v1,
model_name_v2, file_hash_v2)
model_path = fetch_file(model_path,
fname=model_name,
file_hash=file_hash,
cache_subdir='models')
return load_model(model_path), anchors