Hugging Face's logo

STMicroelectronics
/
ssd_mobilenet_v2_fpnlite

Object Detection
Model card Files
xet
Community
FBAGSTM commited on
Commit
77d1a26
·
verified ·
1 Parent(s): 60abbe2

Update Readme ST Model Zoo

Browse files
Files changed (1) hide show
  1. README.md +59 -71
README.md CHANGED
@@ -1,10 +1,3 @@
1
- ---
2
- license: other
3
- license_name: sla0044
4
- license_link: >-
5
- https://github.com/STMicroelectronics/stm32aimodelzoo/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/LICENSE.md
6
- pipeline_tag: object-detection
7
- ---
8
  # SSD MobileNet v2 FPN-lite quantized
9
 
10
  ## **Use case** : `Object detection`
@@ -67,73 +60,68 @@ For an image resolution of NxM and NC classes
67
  Measures are done with default STM32Cube.AI configuration with enabled input / output allocated option.
68
 
69
  ### Reference **NPU** memory footprint based on COCO Person dataset (see Accuracy for details on dataset)
70
- |Model | Dataset | Format | Resolution | Series | Internal RAM (KiB) | External RAM (KiB) | Weights Flash (KiB) | STM32Cube.AI version | STEdgeAI Core version |
71
- |----------|------------------|--------|-------------|------------------|------------------|---------------------|-------|----------------------|-------------------------|
72
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | COCO-Person | Int8 | 192x192x3 | STM32N6 | 606.49 | 0.0 | 1580.53 | 10.0.0 | 2.0.0 |
73
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | COCO-Person | Int8 | 224x224x3 | STM32N6 | 1314.67 | 0.0 | 1607.41 | 10.0.0 | 2.0.0 |
74
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | COCO-Person | Int8 | 256x256x3 | STM32N6 | 1959.06 | 0.0 | 1637.02 | 10.0.0 | 2.0.0 |
75
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | COCO-Person | Int8 | 416x416x3 | STM32N6 | 4570.03 | 0.0 | 1837.8 | 10.0.0 | 2.0.0 |
76
-
77
  ### Reference **NPU** inference time based on COCO Person dataset (see Accuracy for details on dataset)
78
 
79
 
80
- | Model | Dataset | Format | Resolution | Board | Execution Engine | Inference time (ms) | Inf / sec | STM32Cube.AI version | STEdgeAI Core version |
81
- |--------|------------------|--------|-------------|------------------|------------------|---------------------|-------|----------------------|-------------------------|
82
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | COCO-Person | Int8 | 192x192x3 | STM32N6570-DK | NPU/MCU | 14.37 | 69.57 | 10.0.0 | 2.0.0 |
83
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | COCO-Person | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 18.15 | 55.10 | 10.0.0 | 2.0.0 |
84
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | COCO-Person | Int8 | 256x256x3 | STM32N6570-DK | NPU/MCU | 21.73 | 46.03 | 10.0.0 | 2.0.0 |
85
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | COCO-Person | Int8 | 416x416x3 | STM32N6570-DK | NPU/MCU | 114.12 | 8.76 | 10.0.0 | 2.0.0 |
86
-
87
  ### Reference **MCU** memory footprint based on COCO Person dataset (see Accuracy for details on dataset)
88
 
89
 
90
- | Model | Format | Resolution | Series | Activation RAM (KiB) | Runtime RAM (KiB) | Weights Flash (KiB) | Code Flash (KiB) | Total RAM (KiB) | Total Flash (KiB) | STM32Cube.AI version |
91
- |-------------------|--------|------------|---------|----------------|-------------|---------------|------------|-------------|-------------|-----------------------|
92
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | STM32H7 | 521.210.0.0 | 70.26 | 1098.76 | 192.69 | 591.46 | 1291.45 | 10.0.0 | |
93
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | STM32H7 | 956.82 | 70.3 | 1120.63 | 192.84 | 1027.12 | 1313.47 | 10.0.0 |
94
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | STM32H7 | 1238.29 | 70.3 | 1145.24 | 192.81 | 1308.59 | 1338.05 | 10.0.0 |
95
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | STM32H7 | 2869.05 | 70.3 | 1321.02 | 193.23 | 2939.35 | 1514.25 | 10.0.0 |
96
-
97
 
98
  ### Reference **MCU** inference time based on COCO Person dataset (see Accuracy for details on dataset)
99
 
100
 
101
- | Model | Format | Resolution | Board | Execution Engine | Frequency | Inference time (ms) | STM32Cube.AI version |
102
- |-------------------|--------|------------|------------------|------------------|-------------|---------------------|-----------------------|
103
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 511.16 ms | 10.0.0 |
104
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 673.19 ms | 10.0.0 |
105
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 898.32 ms | 10.0.0 |
106
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 2684.93 ms | 10.0.0 |
107
-
108
 
109
  ### Reference **MPU** inference time based on COCO Person dataset (see Accuracy for details on dataset)
110
 
111
  | Model | Format | Resolution | Quantization | Board | Execution Engine | Frequency | Inference time (ms) | %NPU | %GPU | %CPU | X-LINUX-AI version | Framework |
112
  |--------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------|------------|---------------|-------------------|------------------|-----------|---------------------|-------|-------|------|--------------------|-----------------------|
113
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 35.08 ms | 6.20 | 93.80 |0 | v5.1.0 | OpenVX |
114
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 48.92 ms | 6.19 | 93.81 |0 | v5.1.0 | OpenVX |
115
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 40.66 ms | 7.07 | 92.93 |0 | v5.1.0 | OpenVX |
116
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 110.4 ms | 4.47 | 95.53 |0 | v5.1.0 | OpenVX |
117
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 193.70 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
118
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 263.60 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
119
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 339.40 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
120
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 894.00 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
121
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 287.40 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
122
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 383.40 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
123
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 498.90 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
124
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 1348.00 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
125
 
126
 
127
  ### Reference **MPU** inference time based on COCO 80 classes dataset (see Accuracy for details on dataset)
128
 
129
  | Model | Format | Resolution | Quantization | Board | Execution Engine | Frequency | Inference time (ms) | %NPU | %GPU | %CPU | X-LINUX-AI version | Framework |
130
  |--------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------|------------|---------------|-------------------|------------------|-----------|---------------------|-------|-------|------|--------------------|-----------------------|
131
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256_int8.tflite) | Int8 | 256x256x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 100.90 ms | 8.86 | 91.14 |0 | v5.1.0 | OpenVX |
132
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416_int8.tflite) | Int8 | 416x416x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 280.00 ms | 8.68 | 91.32 |0 | v5.1.0 | OpenVX |
133
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256_int8.tflite) | Int8 | 256x256x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 742.90 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
134
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416_int8.tflite) | Int8 | 416x416x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 2000 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
135
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256_int8.tflite) | Int8 | 256x256x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 1112.00 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
136
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416_int8.tflite) | Int8 | 416x416x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 2986 ms | NA | NA |100 | v5.1.0 | TensorFlowLite 2.11.0 |
137
 
138
 
139
 
@@ -144,19 +132,19 @@ Measures are done with default STM32Cube.AI configuration with enabled input / o
144
 
145
  Dataset details: [link](https://cocodataset.org/#download) , License [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) , Quotation[[1]](#1) , Number of classes: 80, Number of images: 118,287
146
 
147
- | Model | Format | Resolution | AP* |
148
- |-------|--------|------------|----------------|
149
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | 40.7 % |
150
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192.h5) | Float | 192x192x3 | 40.8 % |
151
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | 51.1 % |
152
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224.h5) | Float | 224x224x3 | 51.7 % |
153
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | 58.3 % |
154
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256.h5) | Float | 256x256x3 | 58.8 % |
155
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | 61.9 % |
156
- | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416.h5) | Float | 416x416x3 | 62.6 % |
157
 
158
 
159
- \* EVAL_IOU = 0.4, NMS_THRESH = 0.5, SCORE_THRESH =0.001
160
 
161
 
162
  ### AP on COCO 80 classes dataset
@@ -164,14 +152,14 @@ Dataset details: [link](https://cocodataset.org/#download) , License [CC BY 4.0]
164
 
165
  Dataset details: [link](https://cocodataset.org/#download) , License [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) , Quotation[[1]](#1) , Number of classes: 80, Number of images: 118,287
166
 
167
- | Model | Format | Resolution | AP* |
168
- |-------|--------|------------|----------------|
169
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256_int8.tflite) | Int8 | 256x256x3 | 32.2 % |
170
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256.h5) | Float | 256x256x3 | 32.6 % |
171
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416_int8.tflite) | Int8 | 416x416x3 | 32.3 % |
172
- | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416.h5) | Float | 416x416x3 | 34.8 % |
173
 
174
- \* EVAL_IOU = 0.4, NMS_THRESH = 0.5, SCORE_THRESH =0.001
175
 
176
  ## Retraining and Integration in a simple example:
177
 
 
 
 
 
 
 
 
 
1
  # SSD MobileNet v2 FPN-lite quantized
2
 
3
  ## **Use case** : `Object detection`
 
60
  Measures are done with default STM32Cube.AI configuration with enabled input / output allocated option.
61
 
62
  ### Reference **NPU** memory footprint based on COCO Person dataset (see Accuracy for details on dataset)
63
+ | Model | Dataset | Format | Resolution | Series | Internal RAM (KiB) | External RAM (KiB) | Weights Flash (KiB) | STM32Cube.AI version | STEdgeAI Core version |
64
+ |--------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------|----------|--------------|----------|----------------------|----------------------|-----------------------|------------------------|-------------------------|
65
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | COCO-Person | Int8 | 192x192x3 | STM32N6 | 600.08 | 0 | 1480.99 | 10.2.0 | 2.2.0 |
66
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | COCO-Person | Int8 | 224x224x3 | STM32N6 | 1118.31 | 0 | 1507.86 | 10.2.0 | 2.2.0 |
67
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | COCO-Person | Int8 | 256x256x3 | STM32N6 | 1574.94 | 0 | 1537.47 | 10.2.0 | 2.2.0 |
68
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | COCO-Person | Int8 | 416x416x3 | STM32N6 | 2547.84 | 2028 | 1738.25 | 10.2.0 | 2.2.0 |
 
69
  ### Reference **NPU** inference time based on COCO Person dataset (see Accuracy for details on dataset)
70
 
71
 
72
+ | Model | Dataset | Format | Resolution | Board | Execution Engine | Inference time (ms) | Inf / sec | STM32Cube.AI version | STEdgeAI Core version |
73
+ |--------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------|----------|--------------|---------------|--------------------|-----------------------|-------------|------------------------|-------------------------|
74
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | COCO-Person | Int8 | 192x192x3 | STM32N6570-DK | NPU/MCU | 13.5 | 74.07 | 10.2.0 | 2.2.0 |
75
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | COCO-Person | Int8 | 224x224x3 | STM32N6570-DK | NPU/MCU | 17.11 | 58.45 | 10.2.0 | 2.2.0 |
76
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | COCO-Person | Int8 | 256x256x3 | STM32N6570-DK | NPU/MCU | 21.21 | 47.15 | 10.2.0 | 2.2.0 |
77
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | COCO-Person | Int8 | 416x416x3 | STM32N6570-DK | NPU/MCU | 109.65 | 9.12 | 10.2.0 | 2.2.0 |
 
78
  ### Reference **MCU** memory footprint based on COCO Person dataset (see Accuracy for details on dataset)
79
 
80
 
81
+ | Model | Format | Resolution | Series | Activation RAM (KiB) | Runtime RAM (KiB) | Weights Flash (KiB) | Code Flash (KiB) | Total RAM (KiB) | Total Flash (KiB) | STM32Cube.AI version |
82
+ |--------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------|--------------|----------|------------------------|---------------------|-----------------------|--------------------|-------------------|---------------------|------------------------|
83
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | STM32H7 | 956.82 | 70.31 | 1120.63 | 192.19 | 1027.13 | 1312.82 | 10.2.0 |
84
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | STM32H7 | 1238.29 | 70.31 | 1145.24 | 192.16 | 1308.6 | 1337.4 | 10.2.0 |
85
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | STM32H7 | 2869.05 | 70.31 | 1321.02 | 192.58 | 2939.36 | 1513.6 | 10.2.0 |
 
 
86
 
87
  ### Reference **MCU** inference time based on COCO Person dataset (see Accuracy for details on dataset)
88
 
89
 
90
+ | Model | Format | Resolution | Board | Execution Engine | Frequency | Inference time (ms) | STM32Cube.AI version |
91
+ |--------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------|--------------|------------------|--------------------|-------------|-----------------------|------------------------|
92
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 480.95 | 10.2.0 |
93
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 660.75 | 10.2.0 |
94
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 854.94 | 10.2.0 |
95
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | STM32H747I-DISCO | 1 CPU | 400 MHz | 2666.23 | 10.2.0 |
 
96
 
97
  ### Reference **MPU** inference time based on COCO Person dataset (see Accuracy for details on dataset)
98
 
99
  | Model | Format | Resolution | Quantization | Board | Execution Engine | Frequency | Inference time (ms) | %NPU | %GPU | %CPU | X-LINUX-AI version | Framework |
100
  |--------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------|------------|---------------|-------------------|------------------|-----------|---------------------|-------|-------|------|--------------------|-----------------------|
101
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 34.94 ms | 7.32 | 92.68 |0 | v6.1.0 | OpenVX |
102
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 48.44 ms | 6.53 | 93.47 |0 | v6.1.0 | OpenVX |
103
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 40.19 ms | 6.43 | 93.57 |0 | v6.1.0 | OpenVX |
104
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 109.8 ms | 4.96 | 95.04 |0 | v6.1.0 | OpenVX |
105
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 193.48 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
106
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 260.48 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
107
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 337.94 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
108
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 873.31 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
109
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 288.41 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
110
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 392.02 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
111
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 509.83 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
112
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 1339.35 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
113
 
114
 
115
  ### Reference **MPU** inference time based on COCO 80 classes dataset (see Accuracy for details on dataset)
116
 
117
  | Model | Format | Resolution | Quantization | Board | Execution Engine | Frequency | Inference time (ms) | %NPU | %GPU | %CPU | X-LINUX-AI version | Framework |
118
  |--------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------|------------|---------------|-------------------|------------------|-----------|---------------------|-------|-------|------|--------------------|-----------------------|
119
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256_int8.tflite) | Int8 | 256x256x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 98.38 ms | 8.86 | 91.14 |0 | v6.1.0 | OpenVX |
120
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416_int8.tflite) | Int8 | 416x416x3 | per-channel** | STM32MP257F-DK2 | NPU/GPU | 800 MHz | 274.64 ms | 8.38 | 91.62 |0 | v6.1.0 | OpenVX |
121
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256_int8.tflite) | Int8 | 256x256x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 749.00 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
122
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416_int8.tflite) | Int8 | 416x416x3 | per-channel | STM32MP157F-DK2 | 2 CPU | 800 MHz | 1965.36 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
123
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256_int8.tflite) | Int8 | 256x256x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 1124.94 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
124
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416_int8.tflite) | Int8 | 416x416x3 | per-channel | STM32MP135F-DK2 | 1 CPU | 1000 MHz | 2983.66 ms | NA | NA |100 | v6.1.0 | TensorFlowLite 2.18.0 |
125
 
126
 
127
 
 
132
 
133
  Dataset details: [link](https://cocodataset.org/#download) , License [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) , Quotation[[1]](#1) , Number of classes: 80, Number of images: 118,287
134
 
135
+ | Model | Format | Resolution | AP* |
136
+ |-------|--------|------------|-----|
137
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192_int8.tflite) | Int8 | 192x192x3 | 35.0 % |
138
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_192/ssd_mobilenet_v2_fpnlite_035_192.h5) | Float | 192x192x3 | 35.2 % |
139
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224_int8.tflite) | Int8 | 224x224x3 | 45.3 % |
140
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_224/ssd_mobilenet_v2_fpnlite_035_224.h5) | Float | 224x224x3 | 45.5 % |
141
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256_int8.tflite) | Int8 | 256x256x3 | 51.3 % |
142
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_256/ssd_mobilenet_v2_fpnlite_035_256.h5) | Float | 256x256x3 | 51.8 % |
143
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416_int8.tflite) | Int8 | 416x416x3 | 55.4 % |
144
+ | [SSD Mobilenet v2 0.35 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_person/ssd_mobilenet_v2_fpnlite_035_416/ssd_mobilenet_v2_fpnlite_035_416.h5) | Float | 416x416x3 | 56.3 % |
145
 
146
 
147
+ \* EVAL_IOU = 0.5, NMS_THRESH = 0.5, SCORE_THRESH = 0.001, MAX_DETECTIONS = 100
148
 
149
 
150
  ### AP on COCO 80 classes dataset
 
152
 
153
  Dataset details: [link](https://cocodataset.org/#download) , License [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode) , Quotation[[1]](#1) , Number of classes: 80, Number of images: 118,287
154
 
155
+ | Model | Format | Resolution | mAP* |
156
+ |-------|--------|------------|------|
157
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256_int8.tflite) | Int8 | 256x256x3 | 27.1 % |
158
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_256/ssd_mobilenet_v2_fpnlite_100_256.h5) | Float | 256x256x3 | 29.2 % |
159
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416_int8.tflite) | Int8 | 416x416x3 | 28.8 % |
160
+ | [SSD Mobilenet v2 1.0 FPN-lite](https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection/ssd_mobilenet_v2_fpnlite/ST_pretrainedmodel_public_dataset/coco_2017_80_classes/ssd_mobilenet_v2_fpnlite_100_416/ssd_mobilenet_v2_fpnlite_100_416.h5) | Float | 416x416x3 | 31.4 % |
161
 
162
+ \* EVAL_IOU = 0.5, NMS_THRESH = 0.5, SCORE_THRESH = 0.001, MAX_DETECTIONS = 100
163
 
164
  ## Retraining and Integration in a simple example:
165