Hugging Face's logo

STMicroelectronics
/
resnet50v2

Image Classification
Model card Files Community
resnet50v2 / README.md
FBAGSTM's picture
FBAGSTM
Update README.md
f7a97eb verified
preview code
|
raw
history blame contribute delete
8.9 kB
metadata
license: other
license_name: sla0044
license_link: >-
  https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/image_classification/LICENSE.md
pipeline_tag: image-classification

ResNet50 v2

Use case : Image classification

Model description

ResNets family is a well known architecture that uses skip connections to enable stronger gradients in much deeper networks. This variant has 50 layers.

The model is quantized in int8 using tensorflow lite converter.

Network information

Network Information Value
Framework TensorFlow Lite
MParams 25.6 M
Quantization int8
Provenance https://www.tensorflow.org/api_docs/python/tf/keras/applications/ResNet50V2
Paper https://arxiv.org/abs/1603.05027

The models are quantized using tensorflow lite converter.

Network inputs / outputs

For an image resolution of NxM and P classes

Input Shape Description
(1, N, M, 3) Single NxM RGB image with UINT8 values between 0 and 255
Output Shape Description
(1, P) Per-class confidence for P classes in FLOAT32

Recommended platforms

Platform Supported Recommended
STM32L0 [] []
STM32L4 [] []
STM32U5 [] []
STM32H7 [x] []
STM32MP1 [x] []
STM32MP2 [x] [x]
STM32N6 [x] [x]

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 food-101 and ImageNet dataset (see Accuracy for details on dataset)

Model Dataset Format Resolution Series Internal RAM External RAM Weights Flash STM32Cube.AI version STEdgeAI Core version
ResNet50 v2 fft food-101 Int8 224x224x3 STM32N6 10.0.0 2.0.0
ResNet50 v2 ImageNet Int8 224x224x3 STM32N6 10.0.0 2.0.0

Reference NPU inference time on food-101 and ImageNet dataset (see Accuracy for details on dataset)

Model Dataset Format Resolution Board Execution Engine Inference time (ms) Inf / sec STM32Cube.AI version STEdgeAI Core version
ResNet50 v2 fft food-101 Int8 224x224x3 STM32N6570-DK NPU/MCU 10.0.0 2.0.0
ResNet50 v2 ImageNet Int8 224x224x3 STM32N6570-DK NPU/MCU 10.0.0 2.0.0

Reference MCU memory footprint based on Food-101 and ImageNet dataset (see Accuracy for details on dataset)

Model Format Resolution Series Activation RAM Runtime RAM Weights Flash Code Flash Total RAM Total Flash STM32Cube.AI version
ResNet50 v2 fft Int8 224x224x3 STM32H7 2142.07 KiB 41.02 KiB 23240.96 KiB 226.05 KiB 2183.09 KiB 23467.01 KiB 10.0.0
ResNet50 v2 Int8 224x224x3 STM32H7 2142.07 KiB 41.02 KiB 25042.47 KiB 226.05 KiB 2183.09 KiB 25268.52 KiB 10.0.0

Reference MCU inference time based on Food-101 and ImageNet dataset (see Accuracy for details on dataset)

Model Format Resolution Board Execution Engine Frequency Inference time (ms) STM32Cube.AI version
ResNet50 v2 fft Int8 224x224x3 STM32H747I-DISCO 1 CPU 400 MHz 11354.82 ms 10.0.0
ResNet50 v2 Int8 224x224x3 STM32H747I-DISCO 1 CPU 400 MHz 11368.81 ms 10.0.0

Accuracy with Food-101 dataset

Dataset details: link , License -, Quotation[1] , Number of classes: 101 , Number of images: 101 000

Model Format Resolution Top 1 Accuracy
ResNet50 v2 fft Float 224x224x3 71.53 %
ResNet50 v2 fft Int8 224x224x3 70.07 %

Accuracy with ImageNet dataset

Dataset details: link, License: BSD-3-Clause, 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.

model Format Resolution Top 1 Accuracy
ResNet50 v2 Float 224x224x3 66.38 %
ResNet50 v2 Int8 224x224x3 65.99 %

Retraining and Integration in a simple example:

Please refer to the stm32ai-modelzoo-services GitHub here

References

[1] L. Bossard, M. Guillaumin, and L. Van Gool, "Food-101 -- Mining Discriminative Components with Random Forests." European Conference on Computer Vision, 2014.