Source code for tlt.models.image_classification.image_classification_model

#!/usr/bin/env python
# -*- coding: utf-8 -*-
#
# Copyright (c) 2022 Intel Corporation
#
# 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.
#
# SPDX-License-Identifier: Apache-2.0
#

import abc

from tlt.models.model import BaseModel
from tlt.utils.types import FrameworkType, UseCaseType


[docs]class ImageClassificationModel(BaseModel): """ Base class to represent a pretrained model for image classification """
[docs] def __init__(self, image_size, do_fine_tuning: bool, dropout_layer_rate: int, model_name: str, framework: FrameworkType, use_case: UseCaseType): """ Class constructor """ self._image_size = image_size self._do_fine_tuning = do_fine_tuning self._dropout_layer_rate = dropout_layer_rate self._quantization_approach = 'static' BaseModel.__init__(self, model_name, framework, use_case)
@property def image_size(self): """ The fixed image size that the pretrained model expects as input, in pixels with equal width and height """ return self._image_size @property @abc.abstractmethod def num_classes(self): """ The number of output neurons in the model; equal to the number of classes in the dataset """ pass @abc.abstractmethod def predict(self, input_samples): """ Generates predictions for the input samples. The input samples can be a BaseDataset type of object or a numpy array. Returns a numpy array of predictions. """ pass @property def do_fine_tuning(self): """ When True, the weights in all of the model's layers will be trainable. When False, the intermediate layer weights will be frozen, and only the final classification layer will be trainable. """ return self._do_fine_tuning @property def dropout_layer_rate(self): """ The probability of any one node being dropped when a dropout layer is used """ return self._dropout_layer_rate