Updates:

   - Updated How_to_deploy READMEs for Instance Seg, Object Detect and Pose Estimation

Signed-off-by: Guillaume Hortes <[email protected]>

Author: GHORTESSTM

instance_segmentation/deployment/doc/tuto/How_to_deploy_yolov8_instance_segmentation.md

@@ -22,6 +22,26 @@ The STMicroelectronics Ultralytics fork: [https://github.com/stm32-hotspot/ultra
 These models are ready to be deployed and you can go directly to the deployment section.
 The other sections below explain how to start from a model trained with Ultralytics scripts and not quantized.
 
+If you just want to deploy pre-trained and quantized segmentation, you can get them from the STMicroelectronics Ultralytics.
+If you want to train, you can use directly Ultralytics repository at [https://github.com/ultralytics/ultralytics](https://github.com/ultralytics/ultralytics).
+
+## Pre-requisite
+  
+By default, Ultralytics requirements do not install the packages required to export to onnx or tensorflow lite.
+When exporting for the first time, it will either use pre-installed packages or do an auto update installing the latest versions which then causes compatibility issues.
+To ensure compatibility, you need to install (or downgrade) the versions of tensorflow, onnx and onnxruntime following below requirements:
+Use a python 3.9 environment (for the tflite_support package dependency)
+Tensorflow version between 2.8.3 and 2.15.1
+ONNX version between 1.12.0 and 1.15.0
+ONNX runtime version between 1.13 and 1.18.1
+```
+	pip install tensorflow==2.15.1
+	pip install tf_keras==2.15.1
+	pip install onnx==1.15.0
+	pip install onnxruntime==1.18.1
+```
+Other packages can be installed through the auto update procedure.
+
 ## Training a model with Ultralytics scripts
 
 Train the `Yolov8n-seg` model as usual using Ultralytics scripts or start from the pre-trained Yolov8n-seg Pytorch model.

object_detection/deployment/doc/tuto/How_to_deploy_yolov8_yolov5_object_detection.md

@@ -22,6 +22,26 @@ The STMicroelectronics Ultralytics fork: [https://github.com/stm32-hotspot/ultra
 These models are ready to be deployed and you can go directly to the deployment section.
 The other sections below explain how to start from a model trained with Ultralytics scripts and not quantized.
 
+If you just want to deploy pre-trained and quantized segmentation, you can get them from the STMicroelectronics Ultralytics.
+If you want to train, you can use directly Ultralytics repository at [https://github.com/ultralytics/ultralytics](https://github.com/ultralytics/ultralytics).
+
+## Pre-requisite
+  
+By default, Ultralytics requirements do not install the packages required to export to onnx or tensorflow lite.
+When exporting for the first time, it will either use pre-installed packages or do an auto update installing the latest versions which then causes compatibility issues.
+To ensure compatibility, you need to install (or downgrade) the versions of tensorflow, onnx and onnxruntime following below requirements:
+Use a python 3.9 environment (for the tflite_support package dependency)
+Tensorflow version between 2.8.3 and 2.15.1
+ONNX version between 1.12.0 and 1.15.0
+ONNX runtime version between 1.13 and 1.18.1
+```
+	pip install tensorflow==2.15.1
+	pip install tf_keras==2.15.1
+	pip install onnx==1.15.0
+	pip install onnxruntime==1.18.1
+```
+Other packages can be installed through the auto update procedure.
+
 ## Training a model with Ultralytics scripts
 
 Train the `Yolov8n` model as usual using Ultralytics scripts or start from the pre-trained Yolov8n Pytorch model.
@@ -48,14 +68,14 @@ By default the exported models are:
 3. A quantized model per tensor with input / output in integer int8 format: yolov8n_saved_model/yolov8n_integer_quant.tflite.
 4. A quantized model per tensor with input / output in float format: yolov8n_saved_model/yolov8n_full_integer_quant.tflite.
 
-> [!TIPS] It is recommended to use per-channel quantization to better maintain the accuracy, so we recommend to use directly tensorflow lite converter to do the quantization.
+> [!TIP] It is recommended to use per-channel quantization to better maintain the accuracy, so we recommend to use directly tensorflow lite converter to do the quantization.
 
 Start from the generated saved model (1 above) as input for the tensorflow converter. Be sure to used the saved model generated through the export command with int8=True.  
 A script is provided to quantize the model, the yaml file provide the quantization information (see below details).
 
 For deployment the model shall be quantized with input as uint8 and output as float or int8.
 
-> [!Note] Yolov5 
+> [!NOTE] Yolov5 
 
 > The initial version of yolov5n is using a different output shape. For deployment it requires then to add transpose layers compared to the yolov8n. 
 > Ultralytics introduced the yolov5nu version that is aligned with yolov8 output shape.
@@ -205,6 +225,10 @@ Configuration of the post-processing parameters is done through the configuratio
 
 For model with int8 output, the application will detect automatically the zero point and scale to apply for the post processing.  
 
-> [!Note] Yolov5
+> [!NOTE] Yolov5
 
 > According the model used is the yolov5nu for a given resolution, use the same parameters as for yolov8 for the post-processing as they are identical.
+> In the application code, the code enabled by selecting POSTPROCESS_OD_YOLO_V5_UU is deprecated. it corresponds to the older version of yolov5n and not to the yolov5nu.
+> It would require a model with uint8 input and uint8 output.
+> Using the `yolo_v5u` model_type will enable the POSTPROCESS_OD_YOLO_V8_UF or POSTPROCESS_OD_YOLO_V8_UI depending on the input/output format.
+ 

pose_estimation/deployment/doc/tuto/How_to_deploy_yolov8_pose_estimation.md

@@ -22,6 +22,26 @@ The STMicroelectronics Ultralytics fork: [https://github.com/stm32-hotspot/ultra
 These models are ready to be deployed and you can go directly to the deployment section.
 The other sections below explain how to start from a model trained with Ultralytics scripts and not quantized.
 
+If you just want to deploy pre-trained and quantized pose estimation, you can get them from the STMicroelectronics Ultralytics.
+New: if you want to train, you can use now use directly Ultralytics repository at [https://github.com/ultralytics/ultralytics](https://github.com/ultralytics/ultralytics).
+The latest release of Ultralytics fixed the quantization issue on the pose estimation and is now equivalent to ST fork. 
+
+## Pre-requisite
+  
+By default, Ultralytics requirements do not install the packages required to export to onnx or tensorflow lite.
+When exporting for the first time, it will either use pre-installed packages or do an auto update installing the latest versions which then causes compatibility issues.
+To ensure compatibility, you need to install (or downgrade) the versions of tensorflow, onnx and onnxruntime following below requirements:
+Use a python 3.9 environment (for the tflite_support package dependency)
+Tensorflow version between 2.8.3 and 2.15.1
+ONNX version between 1.12.0 and 1.15.0
+ONNX runtime version between 1.13 and 1.18.1
+```
+	pip install tensorflow==2.15.0
+	pip install onnx==1.15.0
+	pip install onnxruntime==1.18.1
+```
+Other packages can be installed through the auto update procedure.
+
 ## Training a model with Ultralytics scripts
 
 Train the `Yolov8n-pose` model as usual using Ultralytics scripts or start from the pre-trained Yolov8n-pose Pytorch model.