Hugging Face's logo

STMicroelectronics
/
miniresnetv2

Audio Classification
Model card Files
xet
Community
FBAGSTM commited on
Commit
14031a1
·
verified ·
1 Parent(s): dd6ede8

Update Readme ST Model Zoo

Browse files
Files changed (1) hide show
  1. README.md +7 -11
README.md CHANGED
@@ -1,7 +1,3 @@
1
- ---
2
- license: apache-2.0
3
- pipeline_tag: audio-classification
4
- ---
5
  # Quantized miniresnetv2
6
 
7
  ## **Use case** : `AED`
@@ -57,8 +53,9 @@ It outputs embedding vectors of size 2048 for the 2 stacks version, and 3548 for
57
 
58
  ## Metrics
59
 
 
60
 
61
- Measures are done with default STM32Cube.AI configuration with enabled input / output allocated option.
62
 
63
 
64
  ### Reference MCU memory footprint based on ESC-10 dataset
@@ -66,8 +63,8 @@ Measures are done with default STM32Cube.AI configuration with enabled input / o
66
 
67
  | Model | Format | Resolution | Series | Activation RAM (KiB) | Runtime RAM (KiB) | Weights Flash (KiB) | Code Flash (KiB) | Total RAM (KiB) | Total Flash (kB) | STM32Cube.AI version |
68
  |-------------------|--------|------------|---------|----------------|-------------|---------------|------------|-------------|-------------|-----------------------|
69
- | [miniresnet v2 1stacks ](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/audio_event_detection/miniresnetv2/ST_pretrainedmodel_public_dataset/esc10/miniresnetv2_1stacks_64x50_tl/miniresnetv2_1stacks_64x50_tl_int8.tflite) | int8 | 64x50x1 | B-U585I-IOT02A | 59.89 | 7.09 | 123.98 | 61.57 | 66.98 | 185.55 | 10.0.0 |
70
- | [miniresnet v2 2stacks ](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/audio_event_detection/miniresnetv2/ST_pretrainedmodel_public_dataset/esc10/miniresnetv2_2stacks_64x50_tl/miniresnetv2_2stacks_64x50_tl_int8.tflite) | int8 | 64x50x1 | B-U585I-IOT02A | 59.89 | 11.28 | 431.98 | 69.86 | 71.17 | 501.84 | 10.0.0 |
71
 
72
 
73
  ### Reference inference time based on ESC-10 dataset
@@ -75,8 +72,8 @@ Measures are done with default STM32Cube.AI configuration with enabled input / o
75
 
76
  | Model | Format | Resolution | Board | Execution Engine | Frequency | Inference time (ms) | STM32Cube.AI version |
77
  |-------------------|--------|------------|------------------|------------------|--------------|-------|-----------------------|
78
- | [miniresnet v2 1stacks ](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/audio_event_detection/miniresnetv2/ST_pretrainedmodel_public_dataset/esc10/miniresnetv2_1stacks_64x50_tl/miniresnetv2_1stacks_64x50_tl_int8.tflite) | int8 | 64x50x1 | B-U585I-IOT02A | 1 CPU | 160 | 188.36 | 10.0.0 |
79
- | [miniresnet v2 2stacks ](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/audio_event_detection/miniresnetv2/ST_pretrainedmodel_public_dataset/esc10/miniresnetv2_2stacks_64x50_tl/miniresnetv2_2stacks_64x50_tl_int8.tflite) | int8 | 64x50x1 | B-U585I-IOT02A | 1 CPU | 160 | 308.81 | 10.0.0 |
80
 
81
 
82
  ### Accuracy with ESC-10 dataset
@@ -96,5 +93,4 @@ The reason this metric is used instead of patch-level accuracy is because patch-
96
 
97
  ## Retraining and Integration in a simple example:
98
 
99
- Please refer to the stm32ai-modelzoo-services GitHub [here](https://github.com/STMicroelectronics/stm32ai-modelzoo-services)
100
-
 
 
 
 
 
1
  # Quantized miniresnetv2
2
 
3
  ## **Use case** : `AED`
 
53
 
54
  ## Metrics
55
 
56
+ * Measures are done with default STM32Cube.AI configuration with enabled input / output allocated option.
57
 
58
+ * `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.
59
 
60
 
61
  ### Reference MCU memory footprint based on ESC-10 dataset
 
63
 
64
  | Model | Format | Resolution | Series | Activation RAM (KiB) | Runtime RAM (KiB) | Weights Flash (KiB) | Code Flash (KiB) | Total RAM (KiB) | Total Flash (kB) | STM32Cube.AI version |
65
  |-------------------|--------|------------|---------|----------------|-------------|---------------|------------|-------------|-------------|-----------------------|
66
+ | [miniresnet v2 1stacks ](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/audio_event_detection/miniresnetv2/ST_pretrainedmodel_public_dataset/esc10/miniresnetv2_1stacks_64x50_tl/miniresnetv2_1stacks_64x50_tl_int8.tflite) | int8 | 64x50x1 | B-U585I-IOT02A | 59.89 | 7.09 | 123.98 | 60.56 9 | 66.98| 184.54 | 10.2.0 |
67
+ | [miniresnet v2 2stacks ](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/audio_event_detection/miniresnetv2/ST_pretrainedmodel_public_dataset/esc10/miniresnetv2_2stacks_64x50_tl/miniresnetv2_2stacks_64x50_tl_int8.tflite) | int8 | 64x50x1 | B-U585I-IOT02A | 59.89 | 11.29 | 431.98 | 68.85 | 71.18 | 500.83 | 10.2.0 |
68
 
69
 
70
  ### Reference inference time based on ESC-10 dataset
 
72
 
73
  | Model | Format | Resolution | Board | Execution Engine | Frequency | Inference time (ms) | STM32Cube.AI version |
74
  |-------------------|--------|------------|------------------|------------------|--------------|-------|-----------------------|
75
+ | [miniresnet v2 1stacks ](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/audio_event_detection/miniresnetv2/ST_pretrainedmodel_public_dataset/esc10/miniresnetv2_1stacks_64x50_tl/miniresnetv2_1stacks_64x50_tl_int8.tflite) | int8 | 64x50x1 | B-U585I-IOT02A | 1 CPU | 160 | 187.21 | 10.2.0 |
76
+ | [miniresnet v2 2stacks ](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/audio_event_detection/miniresnetv2/ST_pretrainedmodel_public_dataset/esc10/miniresnetv2_2stacks_64x50_tl/miniresnetv2_2stacks_64x50_tl_int8.tflite) | int8 | 64x50x1 | B-U585I-IOT02A | 1 CPU | 160 | 307.22 | 10.2.0 |
77
 
78
 
79
  ### Accuracy with ESC-10 dataset
 
93
 
94
  ## Retraining and Integration in a simple example:
95
 
96
+ Please refer to the stm32ai-modelzoo-services GitHub [here](https://github.com/STMicroelectronics/stm32ai-modelzoo-services)