Operator Reference
read_dl_classifier (Operator)
read_dl_classifier — Read a deep-learning-based classifier from a file.
Warning
read_dl_classifier is obsolete and is only provided for
reasons of backward compatibility.
The operator will be removed with HALCON 25.05.
New applications should use the
common CNN-based operator read_dl_model.
Signature
read_dl_classifier( : : FileName : DLClassifierHandle)
Description
The operator read_dl_classifier reads a neural network written
by write_dl_classifier.
As a result, the handle DLClassifierHandle is returned.
HALCON provides pretrained neural networks. These neural networks are good starting points to train a custom classifier for image classification. They have been pretrained on a large image dataset. The provided pretrained neural networks are:
- 'pretrained_dl_classifier_compact.hdl':
-
This neural network is designed to be memory and runtime efficient.
This classifier expects the images to be of the type
real. Additionally, the network is designed for certain image properties. The corresponding values can be retrieved withget_dl_classifier_param. Here we list the default values with which the classifier has been trained:'image_width': 224
'image_height': 224
'image_num_channels': 3
'image_range_min': -127.0
-
'image_range_max': 128.0
This network does not contain any fully connected layer. The network architecture allows changes concerning the image dimensions, but requires a minimum 'image_width' and 'image_height' of 15 pixels.
- 'pretrained_dl_classifier_enhanced.hdl':
-
This neural network has more hidden layers than 'pretrained_dl_classifier_compact.hdl' and is therefore assumed to be better suited for more complex classification tasks. But this comes at the cost of being more time and memory demanding. As a result, e.g., in comparison to the above compact network, the batch size has to be decreased network during the training, see
set_dl_classifier_param.This classifier expects the images to be of the type
real. Additionally, the network is designed for certain image properties. The corresponding values can be retrieved withget_dl_classifier_param. Here we list the default values with which the classifier has been trained:'image_width': 224
'image_height': 224
'image_num_channels': 3
'image_range_min': -127.0
-
'image_range_max': 128.0
The network architecture allows changes concerning the image dimensions, but requires a minimum 'image_width' and 'image_height' of 47 pixels. There is no maximum image size, but large image sizes will increase the memory demand and the runtime significantly. Changing the image size will reinitialize the weights of the fully connected layers and therefore makes a retraining necessary.
- 'pretrained_dl_classifier_resnet18.hdl':
-
As the neural network 'pretrained_dl_classifier_enhanced.hdl', this classifier is suited for more complex tasks. But its structure differs, bringing the advantage of making the training more stable and being internally more robust. Compared to the neural network 'pretrained_dl_classifier_resnet50.hdl' it is less complex and has faster inference times.
This classifier expects the images to be of the type
real. Additionally, the network is designed for certain image properties. The corresponding values can be retrieved withget_dl_classifier_param. Here we list the default values with which the classifier has been trained:'image_width': 224
'image_height': 224
'image_num_channels': 3
'image_range_min': -127.0
-
'image_range_max': 128.0
The network architecture allows changes concerning the image dimensions, but a minimum 'image_width' and 'image_height' of 32 pixels is recommended. There is no maximum image size, but large image sizes will increase the memory demand and the runtime significantly. Despite the fully connected layer a change of the image size does not lead to a reinitialization of the weights.
- 'pretrained_dl_classifier_resnet50.hdl':
-
As the neural network 'pretrained_dl_classifier_enhanced.hdl', this classifier is suited for more complex tasks. But its structure differs, bringing the advantage of making the training more stable and being internally more robust.
This classifier expects the images to be of the type
real. Additionally, the network is designed for certain image properties. The corresponding values can be retrieved withget_dl_classifier_param. Here we list the default values with which the classifier has been trained:'image_width': 224
'image_height': 224
'image_num_channels': 3
'image_range_min': -127.0
-
'image_range_max': 128.0
The network architecture allows changes concerning the image dimensions, but a minimum 'image_width' and 'image_height' of 32 pixels is recommended. There is no maximum image size, but large image sizes will increase the memory demand and the runtime significantly. Despite the fully connected layer a change of the image size does not lead to a reinitialization of the weights.
The values listed above are the default image dimensions and gray value
range for the networks and these are the values with which the classifiers
have been trained.
The network architectures allow different image sizes which can be
set with set_dl_classifier_param. For networks with at least one
fully connected layer such a change makes a retraining necessary.
Networks without fully connected layers are directly applicable to different
image sizes. However, images with a size differing from the size with which
the classifier has been trained are likely to show a reduced classification
accuracy.
The actually configured dimensions can be queried by
get_dl_classifier_param.
Every image that is fed into a network must be present according to
the required dimensions.
To adjust images accordingly, the procedure
preprocess_dl_classifier_images is available.
Typically it is easier, faster and better to retrain a pretrained
classifier for a given classification problem. A pretrained classifier has
already learned good general purpose features. To retrain the network for
a custom problem, the new 'classes' of the classifier have to be
set with set_dl_classifier_param.
The neural network is loaded from the file FileName.
This file is hereby searched in the directory ($HALCONROOT/dl/) as well
as in the currently used directory.
Please note that the runtime specific parameter 'gpu' of the
classifier is not read from file. Instead it is initialized with its default
value (see set_dl_classifier_param).
The default HALCON file extension for deep learning classifiers is
'.hdl'.
For an explanation of the concept of deep-learning-based classification see the introduction of chapter Deep Learning / Classification. The workflow involving this legacy operator is described in the chapter Legacy / DL Classification.
Execution Information
- Multithreading type: reentrant (runs in parallel with non-exclusive operators).
- Multithreading scope: global (may be called from any thread).
- Processed without parallelization.
This operator returns a handle. Note that the state of an instance of this handle type may be changed by specific operators even though the handle is used as an input parameter by those operators.
Parameters
FileName (input_control) filename.read → (string)
File name.
Default: 'pretrained_dl_classifier_compact.hdl'
List of values: 'pretrained_dl_classifier_compact.hdl', 'pretrained_dl_classifier_enhanced.hdl', 'pretrained_dl_classifier_resnet18.hdl', 'pretrained_dl_classifier_resnet50.hdl'
File extension:
.hdl
DLClassifierHandle (output_control) dl_classifier → (handle)
Handle of the deep learning classifier.
Result
If the indicated file is available and the format is correct,
the operator read_dl_classifier returns the value 2 (
H_MSG_TRUE)
.
Otherwise an exception will be raised.
Possible Successors
set_dl_classifier_param,
get_dl_classifier_param,
apply_dl_classifier,
train_dl_classifier_batch
Alternatives
read_dl_model,
read_class_mlp,
read_class_svm
Module
Deep Learning Enhanced