Changes

Thanks to the unstoppable technology progress, nowadays Artificial Intelligence (AI) and specifically Machine Learning (ML) are spreading on low-power, resource-constrained devices as well. In a typical Industrial IoT scenario, this means that [https://en.wikipedia.org/wiki/Edge_computing#Applications edge devices can implement complex inference algorithms that were used to run on the cloud platforms only]. This Technical Note (TN for short) is the first one of a series illustrating how machine learning-based inference test applications are deployed and perform across different embedded platforms, which are eligible for building such intelligent edge devices. The idea is to develop one or more reference applications with the help of well-known frameworks/libraries and to test them on these platforms for comparing performances, resource utilization, development flow, etc. In the following sections, these applications are described in more detail. Each article of this series explores in detail one specific platform or use case. ==Reference application #1: fruit classifier==This application implements a classifier like the one described [[MISC-TN-011:_Running_an_Azure-generated_TensorFlow_Lite_model_on_Mito8M_SoM_using_NXP_eIQ|here]]. There is one notable difference, however, with respect to the linked article. In this case, the model was created from scratch.  === Model creation ===The model was created and trained using Keras, a high-level API of TensorFlow. The following block shows its architecture:<pre>Model: "sequential"_________________________________________________________________Layer (type) Output Shape Param # =================================================================conv2d (Conv2D) (None, 222, 222, 32) 896 _________________________________________________________________activation (Activation) (None, 222, 222, 32) 0 _________________________________________________________________max_pooling2d (MaxPooling2D) (None, 111, 111, 32) 0 _________________________________________________________________conv2d_1 (Conv2D) (None, 109, 109, 32) 9248 _________________________________________________________________activation_1 (Activation) (None, 109, 109, 32) 0 _________________________________________________________________max_pooling2d_1 (MaxPooling2 (None, 54, 54, 32) 0 _________________________________________________________________conv2d_2 (Conv2D) (None, 52, 52, 64) 18496 _________________________________________________________________activation_2 (Activation) (None, 52, 52, 64) 0 _________________________________________________________________max_pooling2d_2 (MaxPooling2 (None, 26, 26, 64) 0 _________________________________________________________________conv2d_3 (Conv2D) (None, 24, 24, 128) 73856 _________________________________________________________________activation_3 (Activation) (None, 24, 24, 128) 0 _________________________________________________________________max_pooling2d_3 (MaxPooling2 (None, 12, 12, 128) 0 _________________________________________________________________flatten (Flatten) (None, 18432) 0 _________________________________________________________________dense (Dense) (None, 256) 4718848 _________________________________________________________________activation_4 (Activation) (None, 256) 0 _________________________________________________________________dropout (Dropout) (None, 256) 0 _________________________________________________________________dense_1 (Dense) (None, 6) 1542 _________________________________________________________________activation_5 (Activation) (None, 6) 0 =================================================================Total params: 4,822,886Trainable params: 4,822,886Non-trainable params: 0</pre> The training was done in the cloud using an AWS EC2 server set up ad hoc. The dataset was created collecting 240 images of 6 different fruits. 75% of the images were used for the training (''training dataset'') and the rest was used for test/validation purposes (''test dataset'', ''validation dataset''). Of course, training the model with a greater number of images would have led to better accuracy, but '''it wouldn't have changed the inference time'''. As the primary goal of the applications built upon this model is to benchmark different platforms, this is acceptable. Obviously, this would not be if this were a real-world application. Several measures were taken to counter the high overfitting tendency due to the small number of images. For instance, new images were synthesized from the existing ones to simulate a larger dataset (''data augmentation''), as shown below:  [[File:Drawio200% augmentation.png|thumb|center|600px|New images synthesized from an existing one. Original image by tookapic from Pixabay.com]]  The following plots show the training history:  [[File:Keras loss history.png|thumb|center|600px|Variation of the loss (blue) and the validation loss (orange) through the epochs during training]]  [[File:Keras acc history.png|thumb|center|600px|Variation of the accuracy (blue) and the validation accuracy (orange) through the epochs during training]] ==Articles in this series==The other articles in this series are:*[[ML-TN-001_-_AI_at_the_edge:_comparison_of_different_embedded_platforms_-_Part_2|Part 2: testing application #1 on Mito8M SoM (NXP i.MX8M)]]*[[ML-TN-001_-_AI_at_the_edge:_comparison_of_different_embedded_platforms_-_Part_3|Part 3: testing application #1 on Xilinx Zynq UltraScale+ MPSoC ZCU104 Evaluation Kit]]*[[ML-TN-001_-_AI_at_the_edge:_comparison_of_different_embedded_platforms_-_Part_4|Part 4: testing application #1 on NXP i.MX8M Plus EVK]]*[[ML-TN-001_-_AI_at_the_edge:_comparison_of_different_embedded_platforms_-_Part_5|Part 5: comparing NXP i.MX8M Plus NPU and Google Coral TPU]]*[[ML-TN-001_-_AI_at_the_edge:_comparison_of_different_embedded_platforms_-_Part_6|Part 6: testing application #1 on Xilinx Zynq UltraScale+ MPSoC ZCU104 Evaluation Kit with PyTorch]]