From eea02bec4710d477bf99a966dea14581606c045c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Kaan=20=C3=87olak?= Date: Wed, 6 Sep 2023 18:44:38 +0300 Subject: [PATCH 1/6] docs(how-to-guides): add training docs for traffic light recognition MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Kaan Çolak --- docs/how-to-guides/integrating-autoware/.pages | 1 + .../training-traffic-light-classifier-module.md | 1 + 2 files changed, 2 insertions(+) create mode 100644 docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md diff --git a/docs/how-to-guides/integrating-autoware/.pages b/docs/how-to-guides/integrating-autoware/.pages index 29eb42aa464..01e6679a68f 100644 --- a/docs/how-to-guides/integrating-autoware/.pages +++ b/docs/how-to-guides/integrating-autoware/.pages @@ -6,3 +6,4 @@ nav: - 4. Creating maps: creating-maps - 5. Launch Autoware: launch-autoware - 6. Tuning parameters and performance: tuning-parameters-and-performance + - 7. Training machine learning models: training-machine-learning-models diff --git a/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md b/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md new file mode 100644 index 00000000000..69b5418f52e --- /dev/null +++ b/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md @@ -0,0 +1 @@ +# Training Traffic Light Classifier Module \ No newline at end of file From 54599fe298d29840290dc11aec7eb6523607e977 Mon Sep 17 00:00:00 2001 From: "pre-commit-ci[bot]" <66853113+pre-commit-ci[bot]@users.noreply.github.com> Date: Wed, 6 Sep 2023 15:52:10 +0000 Subject: [PATCH 2/6] style(pre-commit): autofix --- .../training-traffic-light-classifier-module.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md b/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md index 69b5418f52e..38e5c0426e6 100644 --- a/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md +++ b/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md @@ -1 +1 @@ -# Training Traffic Light Classifier Module \ No newline at end of file +# Training Traffic Light Classifier Module From f91851af348784c22721bb4bdf59b2d2fe2b06e7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Kaan=20=C3=87olak?= Date: Wed, 6 Sep 2023 19:08:02 +0300 Subject: [PATCH 3/6] docs(how-to-guides): refactor directories MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Kaan Çolak --- docs/how-to-guides/.pages | 1 + docs/how-to-guides/index.md | 3 +++ docs/how-to-guides/integrating-autoware/.pages | 3 +-- .../training-traffic-light-classifier-module.md | 1 - .../training-traffic-light-classifier-model.md | 1 + 5 files changed, 6 insertions(+), 3 deletions(-) delete mode 100644 docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md create mode 100644 docs/how-to-guides/training-machine-learning-models/training-traffic-light-classifier-model.md diff --git a/docs/how-to-guides/.pages b/docs/how-to-guides/.pages index e4f75c009f9..c6582ed61fe 100644 --- a/docs/how-to-guides/.pages +++ b/docs/how-to-guides/.pages @@ -1,4 +1,5 @@ nav: - index.md - Integrating Autoware: integrating-autoware + - Training Machine Learning Models: training-machine-learning-models - others diff --git a/docs/how-to-guides/index.md b/docs/how-to-guides/index.md index dc05cd725cb..80a3d81cc7b 100644 --- a/docs/how-to-guides/index.md +++ b/docs/how-to-guides/index.md @@ -4,6 +4,9 @@ - [Overview](integrating-autoware/overview.md) +## Training machine learning models +- [Training Traffic Light Classifier Module](training-machine-learning-models/training-traffic-light-classifier-model.md) + ## Others - [Debug Autoware](others/debug-autoware.md) diff --git a/docs/how-to-guides/integrating-autoware/.pages b/docs/how-to-guides/integrating-autoware/.pages index 01e6679a68f..fe239fd918b 100644 --- a/docs/how-to-guides/integrating-autoware/.pages +++ b/docs/how-to-guides/integrating-autoware/.pages @@ -5,5 +5,4 @@ nav: - 3. Creating vehicle interface: creating-vehicle-interface-package - 4. Creating maps: creating-maps - 5. Launch Autoware: launch-autoware - - 6. Tuning parameters and performance: tuning-parameters-and-performance - - 7. Training machine learning models: training-machine-learning-models + - 6. Tuning parameters and performance: tuning-parameters-and-performance \ No newline at end of file diff --git a/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md b/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md deleted file mode 100644 index 38e5c0426e6..00000000000 --- a/docs/how-to-guides/integrating-autoware/training-machine-learning-models/training-traffic-light-classifier-module.md +++ /dev/null @@ -1 +0,0 @@ -# Training Traffic Light Classifier Module diff --git a/docs/how-to-guides/training-machine-learning-models/training-traffic-light-classifier-model.md b/docs/how-to-guides/training-machine-learning-models/training-traffic-light-classifier-model.md new file mode 100644 index 00000000000..c0a53810978 --- /dev/null +++ b/docs/how-to-guides/training-machine-learning-models/training-traffic-light-classifier-model.md @@ -0,0 +1 @@ +# Training Traffic Light Classifier Model From bfe8fac6314260f6374a833e8d5acc9f66981ced Mon Sep 17 00:00:00 2001 From: "pre-commit-ci[bot]" <66853113+pre-commit-ci[bot]@users.noreply.github.com> Date: Wed, 6 Sep 2023 16:10:21 +0000 Subject: [PATCH 4/6] style(pre-commit): autofix --- docs/how-to-guides/index.md | 1 + docs/how-to-guides/integrating-autoware/.pages | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/docs/how-to-guides/index.md b/docs/how-to-guides/index.md index 80a3d81cc7b..d90d5592120 100644 --- a/docs/how-to-guides/index.md +++ b/docs/how-to-guides/index.md @@ -5,6 +5,7 @@ - [Overview](integrating-autoware/overview.md) ## Training machine learning models + - [Training Traffic Light Classifier Module](training-machine-learning-models/training-traffic-light-classifier-model.md) ## Others diff --git a/docs/how-to-guides/integrating-autoware/.pages b/docs/how-to-guides/integrating-autoware/.pages index fe239fd918b..29eb42aa464 100644 --- a/docs/how-to-guides/integrating-autoware/.pages +++ b/docs/how-to-guides/integrating-autoware/.pages @@ -5,4 +5,4 @@ nav: - 3. Creating vehicle interface: creating-vehicle-interface-package - 4. Creating maps: creating-maps - 5. Launch Autoware: launch-autoware - - 6. Tuning parameters and performance: tuning-parameters-and-performance \ No newline at end of file + - 6. Tuning parameters and performance: tuning-parameters-and-performance From d484ec95cb47366b132b0b5d1e7a26b1521774a6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Kaan=20=C3=87olak?= Date: Tue, 12 Sep 2023 17:58:26 +0300 Subject: [PATCH 5/6] docs(how-to-guides): refactor docs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Kaan Çolak --- docs/how-to-guides/index.md | 2 +- .../training-machine-learning-models/.pages | 2 + .../training-models.md | 37 +++++++++++++++++++ ...training-traffic-light-classifier-model.md | 1 - 4 files changed, 40 insertions(+), 2 deletions(-) create mode 100644 docs/how-to-guides/training-machine-learning-models/.pages create mode 100644 docs/how-to-guides/training-machine-learning-models/training-models.md delete mode 100644 docs/how-to-guides/training-machine-learning-models/training-traffic-light-classifier-model.md diff --git a/docs/how-to-guides/index.md b/docs/how-to-guides/index.md index d90d5592120..bc41dcd152a 100644 --- a/docs/how-to-guides/index.md +++ b/docs/how-to-guides/index.md @@ -6,7 +6,7 @@ ## Training machine learning models -- [Training Traffic Light Classifier Module](training-machine-learning-models/training-traffic-light-classifier-model.md) +- [Model Training and Deploying](training-machine-learning-models/training-models.md) ## Others diff --git a/docs/how-to-guides/training-machine-learning-models/.pages b/docs/how-to-guides/training-machine-learning-models/.pages new file mode 100644 index 00000000000..ef974c96b02 --- /dev/null +++ b/docs/how-to-guides/training-machine-learning-models/.pages @@ -0,0 +1,2 @@ +nav: + - training-models.md \ No newline at end of file diff --git a/docs/how-to-guides/training-machine-learning-models/training-models.md b/docs/how-to-guides/training-machine-learning-models/training-models.md new file mode 100644 index 00000000000..3b8a1c86ba3 --- /dev/null +++ b/docs/how-to-guides/training-machine-learning-models/training-models.md @@ -0,0 +1,37 @@ +# Training Machine Learning Models + +## Overview + +The Autoware offers a comprehensive array of machine learning models, tailored for a wide range of tasks including 2D and 3D object detection, +traffic light recognition and more. These models have been meticulously trained utilizing **[open-mmlab](https://github.com/open-mmlab)**'s extensive repositories. +By leveraging the provided scripts and following the training steps, you have the capability to train these models using your own dataset, +tailoring them to your specific needs. + +Furthermore, you will find the essential conversion scripts to deploy your trained models into Autoware using the **[mmdeploy](https://github.com/open-mmlab/mmdeploy)** repository. + +## Training traffic light classifier model + +The traffic light classifier model within the Autoware has been trained using the **[mmlab/pretrained](https://github.com/open-mmlab/mmpretrain)** repository. +The Autoware offers pretrained models based on EfficientNet-b1 and MobileNet-v2 architectures. +To fine-tune these models, a total of 83,400 images were employed, comprising 58,600 for training, +14,800 for evaluation, and 10,000 for testing. These images represent Japanese traffic lights and were trained using TIER IV's internal dataset. + + +| Name | Input Size | Test Accuracy | +| --------------- | ---------- | ------------- | +| EfficientNet-b1 | 128 x 128 | 99.76% | +| MobileNet-v2 | 224 x 224 | 99.81% | + + +Comprehensive training instructions for the traffic light classifier model are detailed within +the readme file accompanying **"traffic_light_classifier"** package. These instructions will guide you through +the process of training the model using your own dataset. To facilitate your training, we have also provided +an example dataset containing three distinct classes (green, yellow, red), which you can leverage during the training process. + +Detailed instructions for training the traffic light classifier model can be found **[here](https://github.com/autowarefoundation/autoware.universe/blob/main/perception/traffic_light_classifier/README.md)**. + + \ No newline at end of file diff --git a/docs/how-to-guides/training-machine-learning-models/training-traffic-light-classifier-model.md b/docs/how-to-guides/training-machine-learning-models/training-traffic-light-classifier-model.md deleted file mode 100644 index c0a53810978..00000000000 --- a/docs/how-to-guides/training-machine-learning-models/training-traffic-light-classifier-model.md +++ /dev/null @@ -1 +0,0 @@ -# Training Traffic Light Classifier Model From 5865a9c1331caf16572beb0f349e3e4c8e989e87 Mon Sep 17 00:00:00 2001 From: "pre-commit-ci[bot]" <66853113+pre-commit-ci[bot]@users.noreply.github.com> Date: Tue, 12 Sep 2023 14:58:53 +0000 Subject: [PATCH 6/6] style(pre-commit): autofix --- .../training-machine-learning-models/.pages | 2 +- .../training-models.md | 18 ++++++++---------- 2 files changed, 9 insertions(+), 11 deletions(-) diff --git a/docs/how-to-guides/training-machine-learning-models/.pages b/docs/how-to-guides/training-machine-learning-models/.pages index ef974c96b02..f5d7108726f 100644 --- a/docs/how-to-guides/training-machine-learning-models/.pages +++ b/docs/how-to-guides/training-machine-learning-models/.pages @@ -1,2 +1,2 @@ nav: - - training-models.md \ No newline at end of file + - training-models.md diff --git a/docs/how-to-guides/training-machine-learning-models/training-models.md b/docs/how-to-guides/training-machine-learning-models/training-models.md index 3b8a1c86ba3..eba497a1113 100644 --- a/docs/how-to-guides/training-machine-learning-models/training-models.md +++ b/docs/how-to-guides/training-machine-learning-models/training-models.md @@ -2,29 +2,27 @@ ## Overview -The Autoware offers a comprehensive array of machine learning models, tailored for a wide range of tasks including 2D and 3D object detection, -traffic light recognition and more. These models have been meticulously trained utilizing **[open-mmlab](https://github.com/open-mmlab)**'s extensive repositories. -By leveraging the provided scripts and following the training steps, you have the capability to train these models using your own dataset, +The Autoware offers a comprehensive array of machine learning models, tailored for a wide range of tasks including 2D and 3D object detection, +traffic light recognition and more. These models have been meticulously trained utilizing **[open-mmlab](https://github.com/open-mmlab)**'s extensive repositories. +By leveraging the provided scripts and following the training steps, you have the capability to train these models using your own dataset, tailoring them to your specific needs. Furthermore, you will find the essential conversion scripts to deploy your trained models into Autoware using the **[mmdeploy](https://github.com/open-mmlab/mmdeploy)** repository. ## Training traffic light classifier model -The traffic light classifier model within the Autoware has been trained using the **[mmlab/pretrained](https://github.com/open-mmlab/mmpretrain)** repository. -The Autoware offers pretrained models based on EfficientNet-b1 and MobileNet-v2 architectures. +The traffic light classifier model within the Autoware has been trained using the **[mmlab/pretrained](https://github.com/open-mmlab/mmpretrain)** repository. +The Autoware offers pretrained models based on EfficientNet-b1 and MobileNet-v2 architectures. To fine-tune these models, a total of 83,400 images were employed, comprising 58,600 for training, 14,800 for evaluation, and 10,000 for testing. These images represent Japanese traffic lights and were trained using TIER IV's internal dataset. - | Name | Input Size | Test Accuracy | | --------------- | ---------- | ------------- | | EfficientNet-b1 | 128 x 128 | 99.76% | | MobileNet-v2 | 224 x 224 | 99.81% | - -Comprehensive training instructions for the traffic light classifier model are detailed within -the readme file accompanying **"traffic_light_classifier"** package. These instructions will guide you through +Comprehensive training instructions for the traffic light classifier model are detailed within +the readme file accompanying **"traffic_light_classifier"** package. These instructions will guide you through the process of training the model using your own dataset. To facilitate your training, we have also provided an example dataset containing three distinct classes (green, yellow, red), which you can leverage during the training process. @@ -34,4 +32,4 @@ Detailed instructions for training the traffic light classifier model can be fou Training traffic light detection model and lidar CenterPoint model will be added there. ---> \ No newline at end of file +-->