mobilenetv1

MobileNet v1

Use case : Image classification

Model description

MobileNet is a well known architecture that can be used in multiple use cases. Input size and width factor called alpha are parameters to be adapted to various use cases complexity. The alpha parameter is used to increase or decrease the number of filters in each layer, allowing also to reduce the number of multiply-adds and then the inference time.

The original paper demonstrates the performance of MobileNet models using alpha values of 1.0, 0.75, 0.5 and 0.25.

(source: https://keras.io/api/applications/mobilenet/)

The model is quantized in int8 using tensorflow lite converter.

Network information

The models are quantized using tensorflow lite converter.

Network inputs / outputs

For an image resolution of NxM and P classes

Recommended platforms

Performances

Metrics

• Measures are done with default STM32Cube.AI configuration with enabled input / output allocated option.
• tfs stands for "training from scratch", meaning that the model weights were randomly initialized before training.
• tl stands for "transfer learning", meaning that the model backbone weights were initialized from a pre-trained model, then only the last layer was unfrozen during the training.
• fft stands for "full fine-tuning", meaning that the full model weights were initialized from a transfer learning pre-trained model, and all the layers were unfrozen during the training.

Reference NPU memory footprint on food101 and imagenet dataset (see Accuracy for details on dataset)

Reference NPU inference time on food101 and imagenet dataset (see Accuracy for details on dataset)

Reference MCU memory footprint based on Flowers dataset and imagenet dataset (see Accuracy for details on dataset)

Reference MCU inference time based on Flowers dataset and imagenet dataset (see Accuracy for details on dataset)

Reference MPU inference time based on Flowers dataset (see Accuracy for details on dataset)

** To get the most out of MP25 NPU hardware acceleration, please use per-tensor quantization

** Note: On STM32MP2 devices, per-channel quantized models are internally converted to per-tensor quantization by the compiler using an entropy-based method. This may introduce a slight loss in accuracy compared to the original per-channel models.

Accuracy with Flowers dataset

Dataset details: link , License CC BY 2.0 , Quotation[1] , Number of classes: 5, Number of images: 3 670

Accuracy with Plant-village dataset

Dataset details: link , License CC0 1.0, Quotation[2] , Number of classes: 39, Number of images: 61 486

Accuracy with Food-101 dataset

Dataset details: link, Quotation[3] , Number of classes: 101 , Number of images: 101 000

Accuracy with imagenet dataset

Dataset details: link, Quotation[4]. Number of classes: 1000. To perform the quantization, we calibrated the activations with a random subset of the training set. For the sake of simplicity, the accuracy reported here was estimated on the 50000 labelled images of the validation set.

Retraining and Integration in a simple example:

Please refer to the stm32ai-modelzoo-services GitHub here

References

<a id="1">[1]</a> "Tf_flowers : tensorflow datasets," TensorFlow. [Online]. Available: https://www.tensorflow.org/datasets/catalog/tf_flowers.

<a id="2">[2]</a> J, ARUN PANDIAN; GOPAL, GEETHARAMANI (2019), "Data for: Identification of Plant Leaf Diseases Using a 9-layer Deep Convolutional Neural Network", Mendeley Data, V1, doi: 10.17632/tywbtsjrjv.1

<a id="3">[3]</a> L. Bossard, M. Guillaumin, and L. Van Gool, "Food-101 -- Mining Discriminative Components with Random Forests." European Conference on Computer Vision, 2014.