update-to-version-2.8.0

README.rst

@@ -22,14 +22,14 @@ Hailo Model Zoo
    :height: 20
 
 
-.. |compiler| image:: https://img.shields.io/badge/Hailo%20Dataflow%20Compiler-3.23.0-brightgreen.svg
+.. |compiler| image:: https://img.shields.io/badge/Hailo%20Dataflow%20Compiler-3.24.0-brightgreen.svg
    :target: https://hailo.ai/contact-us/
    :alt: Hailo Dataflow Compiler
    :width: 180
    :height: 20
 
 
-.. |runtime| image:: https://img.shields.io/badge/HailoRT%20(optional)-4.13.0-brightgreen.svg
+.. |runtime| image:: https://img.shields.io/badge/HailoRT%20(optional)-4.14.0-brightgreen.svg
    :target: https://hailo.ai/contact-us/
    :alt: HailoRT
    :width: 170

docs/CHANGELOG.rst

@@ -1,6 +1,41 @@
 Changelog
 =========
 
+**v2.8**
+
+* Update to use Dataflow Compiler v3.24.0 (`developer-zone <https://hailo.ai/developer-zone/>`_)
+* Updated to use HailoRT 4.14.0 (`developer-zone <https://hailo.ai/developer-zone/>`_)
+* The Hailo Model Zoo now supports the following vision transformers models:
+
+  * vit_tiny / vit_small / vit_base - encoder based transformer with batchnorm for classification
+  * detr_resnet_v1_18_bn - encoder/decoder transformer for object detection
+  * clip_resnet_50 - Contrastive Language-Image Pre-Training for zero-shot classification
+  * yolov5s_c3tr - object detection model with a MHSA block
+
+* Using HailoRT-pp for postprocessing of the following variants:
+
+  * yolov5
+  * yolox
+  * ssd
+  * efficientdet
+  * yolov7
+
+* New Models:
+
+  * repvgg_a1 / repvgg_a2 - classification
+  * yolov8_seg: yolov8n_seg / yolov8s_seg / yolov8m_seg - instance segmentation
+  * yolov6n_0.2.1 - object detecion
+  * zero_dce - low-light enhancement 
+
+* New retraining dockers for:
+
+  * yolov8
+  * yolov8_seg
+
+* Enable compilation for hailo15h device
+* Enable evaluation of models with RGBX / NV12 input format
+* Bug fixes
+
 **v2.7**
 
 * Update to use Dataflow Compiler v3.23.0 (`developer-zone <https://hailo.ai/developer-zone/>`_)

docs/DATA.rst

@@ -34,6 +34,8 @@ We recommend to define the data directory path yourself, by setting the ``HMZ_DA
   * `CelebA`_
   * `LFW`_
   * `BSD100`_
+  * `CIFAR100`_
+  * `LOL`_
 
 .. _ImageNet:
 
@@ -728,3 +730,71 @@ Manual Download (Optional)
 
       python hailo_model_zoo/datasets/create_bsd100_tfrecord.py val --lr /path/to/LRbicx4 --hr /path/to/GTmod12
       python hailo_model_zoo/datasets/create_bsd100_tfrecord.py calib --lr /path/to/LRbicx4 --hr /path/to/GTmod12
+
+
+.. _CIFAR100:
+
+CIFAR100
+------
+
+To evaluate/optimize/compile the CLIP models of the 
+Hailo Model Zoo you should generate the CIFAR100 TFRecord files.
+
+Run the creation scripts:
+
+.. code-block::
+
+    python hailo_model_zoo/datasets/create_cifar100_tfrecord.py val
+    python hailo_model_zoo/datasets/create_cifar100_tfrecord.py calib
+
+
+.. _LOL:
+
+LOL
+------
+
+To evaluate/optimize/compile the low light enhancement models of the
+Hailo Model Zoo you should generate the LOL TFRecord files.
+
+Run the creation scripts:
+
+.. code-block::
+
+    python hailo_model_zoo/datasets/create_lol_tfrecord.py val
+    python hailo_model_zoo/datasets/create_lol_tfrecord.py calib
+
+Manual Download (Optional)
+^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+#. Download the LOL dataset from `here <"https://drive.google.com/uc?export=download&id=157bjO1_cFuSd0HWDUuAmcHRJDVyWpOxB&authuser=0">`_ and extract.
+   The expected dataset structure:
+
+   .. code-block::
+
+      lol_dataset
+      |_ eval15
+         |_ high
+         |   |_ 111.png
+         |   |_ 146.png
+         |   |_ ...
+         |_ low
+         |   |_ 111.png
+         |   |_ 146.png
+         |   |_ ...
+      |_ our485
+         |_ high
+         |   |_ 100.png
+         |   |_ 101.png
+         |   |_ ...
+         |_ low
+         |   |_ 100.png
+         |   |_ 101.png
+         |   |_ ...
+
+
+#. Run the scripts:
+
+   .. code-block::
+
+      python hailo_model_zoo/datasets/create_lol_tfrecord.py val --ll /path/to/val/lowlight/images --lle /path/to/val/highlight/images
+      python hailo_model_zoo/datasets/create_lol_tfrecord.py calib --ll /path/to/train/lowlight/images --lle /path/to/train/highlight/images

docs/GETTING_STARTED.rst

@@ -9,8 +9,8 @@ System Requirements
 
 * Ubuntu 20.04/22.04, 64 bit (supported also on Windows, under WSL2)
 * Python 3.8/3.9/3.10, including ``pip`` and ``virtualenv``
-* Hailo Dataflow Compiler v3.23.0 (Obtain from `hailo.ai <http://hailo.ai>`_\ )
-* HailoRT 4.13.0 (Obtain from `hailo.ai <http://hailo.ai>`_\ ) - required only for inference on Hailo-8.
+* Hailo Dataflow Compiler v3.24.0 (Obtain from `hailo.ai <http://hailo.ai>`_\ )
+* HailoRT 4.14.0 (Obtain from `hailo.ai <http://hailo.ai>`_\ ) - required only for inference on Hailo-8.
 * The Hailo Model Zoo supports Hailo-8 connected via PCIe only.
 * Nvidia’s Pascal/Turing/Ampere GPU architecture (such as Titan X Pascal, GTX 1080 Ti, RTX 2080 Ti, or RTX A4000)
 * GPU driver version 470
@@ -109,6 +109,12 @@ The pre-trained models are stored on AWS S3 and will be downloaded automatically
 
    hailomz parse <model_name>
 
+* The default compilation target is Hailo-8. To compile for different architecture (Hailo-15H for example), use ``--hw_arch hailo15h`` as CLI argument:
+
+.. code-block::
+
+   hailomz parse <model_name> --hw-arch hailo15h
+
 Profiling
 ---------
 
@@ -167,17 +173,23 @@ To run the Hailo compiler and generate the Hailo Executable Format (HEF) file:
 
    hailomz compile <model_name>
 
+By default the compilation target is Hailo-8. To compile for a different architecture use ``--hw-arch`` command line argument:
+
+.. code-block::
+
+   hailomz compile <model_name> --hw-arch hailo15h
+
 To generate the HEF starting from a previously generated HAR file:
 
 .. code-block::
 
-   hailomz compile <model_name> --har /path/to/model.har
+   hailomz compile <model_name> --har /path/to/model.har --hw-arch <hailo8|hailo15h>
 
 In order to achieve highest performance, one could use the performance flag:
 
 .. code-block::
 
-    hailomz optimize <model_name> --performance
+    hailomz optimize <model_name> --performance --hw-arch <hailo8|hailo15h>
 
 The flag will be ignored on models that do not support this feature.
 The default and performance model scripts are located on `hailo_model_zoo/cfg/alls/`
@@ -230,6 +242,8 @@ To explore other options (for example: changing the default batch-size) use:
 
    hailomz eval --help
 
+* Currently MZ evaluation can be done only on hailo8
+
 Visualization
 -------------
 

docs/HAILO_MODELS.rst

@@ -7,7 +7,6 @@ Each model is accompanied with its own README, retraining docker and retraining
 
 
 * FLOPs in the table are counted as MAC operations.
-* All models were compiled using Hailo Dataflow Compiler v3.23.0
 * Supported tasks:
 
   * `Object Detection`_