Skip to content

GitLab

Unverified Commit 3d4c14e6 authored by Adrià Arrufat's avatar Adrià Arrufat Committed by GitHub
Browse files

Add input_rgb_image_pair layer and new net visitors documentation (#2497) 

* Add input_rgb_image layer documentation

* fix typo

* Add index entries for missing network visitors

* update Barlow Twins loss links

* add missing yolo rect docs, links to yolo paper and better Barlow Twins formatting
parent 21651f49
Show whitespace changes
Inline Side-by-side
Showing with 77 additions and 7 deletions
docs/docs/imaging.xml
View file @ 3d4c14e6
......@@ -98,6 +98,7 @@
<item>evaluate_detectors</item>
<item>full_object_detection</item>
<item>mmod_rect</item>
<item>yolo_rect</item>
<item>scan_image</item>
<item>scan_image_movable_parts</item>
<item>find_points_above_thresh</item>
......@@ -2551,6 +2552,19 @@
</component>
<!-- ************************************************************************* -->
<component>
<name>yolo_rect</name>
<file>dlib/image_processing.h</file>
<spec_file link="true">dlib/image_processing/full_object_detection_abstract.h</spec_file>
<description>
This is a simple struct that is used to give training data and receive detections
from the <a href="ml.html#loss_yolo_">YOLO loss layer for object detection</a>.
</description>
</component>
<!-- ************************************************************************* -->
<component>
......
docs/docs/ml.xml
View file @ 3d4c14e6
......@@ -126,6 +126,7 @@ Davis E. King. <a href="http://jmlr.csail.mit.edu/papers/volume10/king09a/king09
<item>input_rgb_image</item>
<item>input_rgb_image_sized</item>
<item>input_rgb_image_pyramid</item>
<item>input_rgb_image_pair</item>
<item>input_grayscale_image_pyramid</item>
<item>
<name>EXAMPLE_INPUT_LAYER</name>
......@@ -355,7 +356,7 @@ Davis E. King. <a href="http://jmlr.csail.mit.edu/papers/volume10/king09a/king09
</item>
<item>
<name>loss_barlow_twins</name>
<link>dlib/dnn/loss_abstract.h.html#loss_barlow_twins_</link>
<link>#loss_barlow_twins_</link>
</item>
</sub>
</item>
......@@ -817,6 +818,28 @@ Davis E. King. <a href="http://jmlr.csail.mit.edu/papers/volume10/king09a/king09
</description>
</component>
<!-- ************************************************************************* -->
<component>
<name>input_rgb_image_pair</name>
<file>dlib/dnn.h</file>
<spec_file link="true">dlib/dnn/input_abstract.h</spec_file>
<description>
This is a simple input layer type for use in a deep neural network
which takes a pair of RGB images as input and loads it into a network.
It is useful when you want to input image pairs that are related to each other,
for instance, they are different distorted views of the same original image.
This input layer is meant to be used with a loss layer such as the
<a href="#loss_barlow_twins_">Barlow Twins loss layer</a>. You can also
convert between <a href="#input_rgb_image">input_rgb_image</a> and
this input layer by copy construction or assignment.
</description>
<examples>
<example>dnn_self_supervised_learning_ex.cpp.html</example>
</examples>
</component>
<!-- ************************************************************************* -->
<component>
......@@ -858,7 +881,7 @@ Davis E. King. <a href="http://jmlr.csail.mit.edu/papers/volume10/king09a/king09
This object is a <a href="dlib/dnn/loss_abstract.h.html#EXAMPLE_LOSS_LAYER_">loss layer</a>
for a deep neural network. In particular, it implements the YOLO detection loss defined in
the paper:
<blockquote>YOLOv3: An Incremental Improvement by Joseph Redmon and Ali Farhadi.</blockquote>
<blockquote><a href="https://arxiv.org/abs/1804.02767">YOLOv3: An Incremental Improvement</a> by Joseph Redmon and Ali Farhadi.</blockquote>
This means you use this loss if you want to detect the locations of objects
in images.
......@@ -916,6 +939,32 @@ Davis E. King. <a href="http://jmlr.csail.mit.edu/papers/volume10/king09a/king09
</description>
</component>
<!-- ************************************************************************* -->
<component>
<name>loss_barlow_twins_</name>
<file>dlib/dnn.h</file>
<spec_file link="true">dlib/dnn/loss_abstract.h</spec_file>
<description>
This object is a <a href="dlib/dnn/loss_abstract.h.html#EXAMPLE_LOSS_LAYER_">loss layer</a>
for a deep neural network. In particular, it implements the Barlow Twins loss layer presented
in the paper:
<blockquote>
<a href="https://arxiv.org/abs/2103.03230">Barlow Twins: Self-Supervised Learning
via Redundancy Reduction</a>, by Jure Zbontar, Li Jing, Ishan Misra, Yann LeCun, St&#233;phane Deny.
</blockquote>
This means you use this loss to learn useful representations from data that has no label
information. Useful representations mean that can be used to train another downstream task,
such as classification. In particular, this loss function applies the redundancy reduction
principle to the representations learned by the network it sits on top of.
</description>
<examples>
<example>dnn_self_supervised_learning_ex.cpp.html</example>
</examples>
</component>
<!-- ************************************************************************* -->
<component>
......
docs/docs/term_index.xml
View file @ 3d4c14e6
......@@ -39,6 +39,9 @@
<term file="dlib/dnn/utilities_abstract.h.html" name="randomize_parameters" include="dlib/dnn.h"/>
<term file="dlib/dnn/utilities_abstract.h.html" name="input_tensor_to_output_tensor" include="dlib/dnn.h"/>
<term file="dlib/dnn/utilities_abstract.h.html" name="output_tensor_to_input_tensor" include="dlib/dnn.h"/>
<term file="dlib/dnn/utilities_abstract.h.html" name="count_parameters" include="dlib/dnn.h"/>
<term file="dlib/dnn/utilities_abstract.h.html" name="set_all_learning_rate_multipliers" include="dlib/dnn.h"/>
<term file="dlib/dnn/utilities_abstract.h.html" name="set_learning_rate_multipliers_range" include="dlib/dnn.h"/>
<term file="dlib/dnn/core_abstract.h.html" name="tuple_head" include="dlib/dnn.h"/>
<term file="dlib/dnn/core_abstract.h.html" name="tuple_tail" include="dlib/dnn.h"/>
<term file="dlib/dnn/core_abstract.h.html" name="get_learning_rate_multiplier" include="dlib/dnn.h"/>
......@@ -101,6 +104,8 @@
<term file="dlib/dnn/layers_abstract.h.html" name="CONV_MODE" include="dlib/dnn.h"/>
<term file="dlib/dnn/layers_abstract.h.html" name="FC_MODE" include="dlib/dnn.h"/>
<term file="dlib/dnn/layers_abstract.h.html" name="set_all_bn_running_stats_window_sizes" include="dlib/dnn.h"/>
<term file="dlib/dnn/layers_abstract.h.html" name="disable_duplicative_biases" include="dlib/dnn.h"/>
<term file="dlib/dnn/layers_abstract.h.html" name="fuse_layers" include="dlib/dnn.h"/>
<term file="dlib/cuda/tensor_abstract.h.html" name="tensor" include="dlib/cuda/tensor.h"/>
<term file="dlib/cuda/tensor_abstract.h.html" name="resizable_tensor" include="dlib/cuda/tensor.h"/>
<term file="dlib/cuda/tensor_abstract.h.html" name="alias_tensor_instance" include="dlib/cuda/tensor.h"/>
......@@ -123,6 +128,7 @@
<term file="ml.html" name="input_rgb_image" include="dlib/dnn.h"/>
<term file="ml.html" name="input_rgb_image_sized" include="dlib/dnn.h"/>
<term file="ml.html" name="input_rgb_image_pyramid" include="dlib/dnn.h"/>
<term file="ml.html" name="input_rgb_image_pair" include="dlib/dnn.h"/>
<term file="ml.html" name="input_grayscale_image_pyramid" include="dlib/dnn.h"/>
<term file="ml.html" name="dnn_trainer" include="dlib/dnn.h"/>
......@@ -149,7 +155,7 @@
<term file="dlib/dnn/loss_abstract.h.html" name="mmod_options" include="dlib/dnn.h"/>
<term file="ml.html" name="loss_yolo_" include="dlib/dnn.h"/>
<term file="dlib/dnn/loss_abstract.h.html" name="yolo_options" include="dlib/dnn.h"/>
<term file="dlib/dnn/loss_abstract.h.html" name="loss_barlow_twins_" include="dlib/dnn.h"/>
<term file="ml.html" name="loss_barlow_twins_" include="dlib/dnn.h"/>
<term file="dlib/dnn/solvers_abstract.h.html" name="EXAMPLE_SOLVER" include="dlib/dnn.h"/>
<term file="dlib/dnn/solvers_abstract.h.html" name="sgd" include="dlib/dnn.h"/>
......@@ -1649,6 +1655,7 @@
<term file="dlib/image_processing/object_detector_abstract.h.html" name="full_detection" include="dlib/image_processing.h"/>
<term file="imaging.html" name="full_object_detection" include="dlib/image_processing.h"/>
<term file="imaging.html" name="mmod_rect" include="dlib/image_processing.h"/>
<term file="imaging.html" name="yolo_rect" include="dlib/image_processing.h"/>
<term file="dlib/image_processing/full_object_detection_abstract.h.html" name="all_parts_in_rect" include="dlib/image_processing.h"/>
<term file="imaging.html" name="scan_image_movable_parts" include="dlib/image_processing.h"/>
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment