Hugging Face's logo

qualcomm
/
FFNet-122NS-LowRes

Image Segmentation
PyTorch
android
Model card Files
xet
Community
FFNet-122NS-LowRes
File size: 10,576 Bytes 
28f7c59
 
e7dbfaa
28f7c59
 
a4390f4
28f7c59
 
 
1d5c1e0
28f7c59
fe54f70
956b84a
28f7c59
 
fe54f70
 
 
 
 
 
 
 
 
 
 
 
 
 
7d21d35
 
 
 
 
 
fe54f70
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c655677
28f7c59
6a4bb53
 
c655677
28f7c59
 
 
 
 
875bb11
28f7c59
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
 
---
library_name: pytorch
license: other
tags:
- android
pipeline_tag: image-segmentation

---

![](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_122ns_lowres/web-assets/model_demo.png)

# FFNet-122NS-LowRes: Optimized for Qualcomm Devices

FFNet-122NS-LowRes is a "fuss-free network" that segments street scene images with per-pixel classes like road, sidewalk, and pedestrian. Trained on the Cityscapes dataset.

This is based on the implementation of FFNet-122NS-LowRes found [here](https://github.com/Qualcomm-AI-research/FFNet).
This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/ffnet_122ns_lowres) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).

Qualcomm AI Hub Models uses [Qualcomm AI Hub Workbench](https://workbench.aihub.qualcomm.com) to compile, profile, and evaluate this model. [Sign up](https://myaccount.qualcomm.com/signup) to run these models on a hosted Qualcomm® device.

## Getting Started
There are two ways to deploy this model on your device:

### Option 1: Download Pre-Exported Models

Below are pre-exported model assets ready for deployment.

| Runtime | Precision | Chipset | SDK Versions | Download |
|---|---|---|---|---|
| ONNX | float | Universal | QAIRT 2.37, ONNX Runtime 1.23.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_122ns_lowres/releases/v0.46.0/ffnet_122ns_lowres-onnx-float.zip)
| ONNX | w8a8 | Universal | QAIRT 2.37, ONNX Runtime 1.23.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_122ns_lowres/releases/v0.46.0/ffnet_122ns_lowres-onnx-w8a8.zip)
| QNN_DLC | float | Universal | QAIRT 2.42 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_122ns_lowres/releases/v0.46.0/ffnet_122ns_lowres-qnn_dlc-float.zip)
| QNN_DLC | w8a8 | Universal | QAIRT 2.42 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_122ns_lowres/releases/v0.46.0/ffnet_122ns_lowres-qnn_dlc-w8a8.zip)
| TFLITE | float | Universal | QAIRT 2.42, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_122ns_lowres/releases/v0.46.0/ffnet_122ns_lowres-tflite-float.zip)
| TFLITE | w8a8 | Universal | QAIRT 2.42, TFLite 2.17.0 | [Download](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/models/ffnet_122ns_lowres/releases/v0.46.0/ffnet_122ns_lowres-tflite-w8a8.zip)

For more device-specific assets and performance metrics, visit **[FFNet-122NS-LowRes on Qualcomm® AI Hub](https://aihub.qualcomm.com/models/ffnet_122ns_lowres)**.


### Option 2: Export with Custom Configurations

Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/ffnet_122ns_lowres) Python library to compile and export the model with your own:
- Custom weights (e.g., fine-tuned checkpoints)
- Custom input shapes
- Target device and runtime configurations

This option is ideal if you need to customize the model beyond the default configuration provided here.

See our repository for [FFNet-122NS-LowRes on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/ffnet_122ns_lowres) for usage instructions.

## Model Details

**Model Type:** Model_use_case.semantic_segmentation

**Model Stats:**
- Model checkpoint: ffnet122NS_CCC_cityscapes_state_dict_quarts_pre_down
- Input resolution: 1024x512
- Number of output classes: 19
- Number of parameters: 32.1M
- Model size (float): 123 MB
- Model size (w8a8): 31.3 MB

## Performance Summary
| Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
|---|---|---|---|---|---|---
| FFNet-122NS-LowRes | ONNX | float | Snapdragon® X Elite | 7.035 ms | 56 - 56 MB | NPU
| FFNet-122NS-LowRes | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 5.043 ms | 3 - 166 MB | NPU
| FFNet-122NS-LowRes | ONNX | float | Qualcomm® QCS8550 (Proxy) | 7.045 ms | 0 - 59 MB | NPU
| FFNet-122NS-LowRes | ONNX | float | Qualcomm® QCS9075 | 10.371 ms | 6 - 15 MB | NPU
| FFNet-122NS-LowRes | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.258 ms | 3 - 126 MB | NPU
| FFNet-122NS-LowRes | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 3.671 ms | 0 - 125 MB | NPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Snapdragon® X Elite | 2.793 ms | 30 - 30 MB | NPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Snapdragon® 8 Gen 3 Mobile | 2.036 ms | 0 - 220 MB | NPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Qualcomm® QCS6490 | 91.933 ms | 54 - 153 MB | CPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Qualcomm® QCS8550 (Proxy) | 2.86 ms | 0 - 37 MB | NPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Qualcomm® QCS9075 | 3.305 ms | 1 - 4 MB | NPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Qualcomm® QCM6690 | 84.154 ms | 60 - 73 MB | CPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.716 ms | 0 - 157 MB | NPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Snapdragon® 7 Gen 4 Mobile | 83.743 ms | 61 - 75 MB | CPU
| FFNet-122NS-LowRes | ONNX | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 1.515 ms | 0 - 157 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Snapdragon® X Elite | 13.589 ms | 6 - 6 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 8.817 ms | 0 - 201 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 39.547 ms | 0 - 160 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 13.343 ms | 6 - 9 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Qualcomm® SA8775P | 16.002 ms | 0 - 162 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Qualcomm® QCS9075 | 16.888 ms | 6 - 14 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 29.955 ms | 6 - 197 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Qualcomm® SA7255P | 39.547 ms | 0 - 160 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Qualcomm® SA8295P | 17.669 ms | 0 - 152 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.99 ms | 0 - 163 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.654 ms | 6 - 172 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Snapdragon® X Elite | 4.839 ms | 2 - 2 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Snapdragon® 8 Gen 3 Mobile | 3.191 ms | 2 - 130 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® QCS6490 | 13.871 ms | 4 - 7 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® QCS8275 (Proxy) | 9.288 ms | 1 - 76 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® QCS8550 (Proxy) | 4.5 ms | 2 - 3 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® SA8775P | 21.084 ms | 1 - 76 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® QCS9075 | 5.871 ms | 2 - 5 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® QCM6690 | 27.238 ms | 2 - 199 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® QCS8450 (Proxy) | 6.928 ms | 2 - 126 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® SA7255P | 9.288 ms | 1 - 76 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Qualcomm® SA8295P | 6.037 ms | 2 - 74 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 2.194 ms | 2 - 79 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Snapdragon® 7 Gen 4 Mobile | 5.687 ms | 2 - 193 MB | NPU
| FFNet-122NS-LowRes | QNN_DLC | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 1.803 ms | 2 - 80 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 8.64 ms | 1 - 280 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 39.759 ms | 1 - 194 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 12.666 ms | 1 - 3 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Qualcomm® SA8775P | 16.045 ms | 1 - 194 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Qualcomm® QCS9075 | 16.8 ms | 0 - 70 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 30.193 ms | 1 - 271 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Qualcomm® SA7255P | 39.759 ms | 1 - 194 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Qualcomm® SA8295P | 17.656 ms | 1 - 191 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 6.882 ms | 0 - 199 MB | NPU
| FFNet-122NS-LowRes | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 5.061 ms | 1 - 199 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Snapdragon® 8 Gen 3 Mobile | 1.872 ms | 0 - 136 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® QCS6490 | 9.208 ms | 0 - 35 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® QCS8275 (Proxy) | 5.961 ms | 0 - 71 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® QCS8550 (Proxy) | 2.619 ms | 0 - 35 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® SA8775P | 3.086 ms | 0 - 74 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® QCS9075 | 3.041 ms | 0 - 35 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® QCM6690 | 23.267 ms | 0 - 198 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® QCS8450 (Proxy) | 3.285 ms | 0 - 129 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® SA7255P | 5.961 ms | 0 - 71 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Qualcomm® SA8295P | 3.868 ms | 0 - 69 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Snapdragon® 8 Elite For Galaxy Mobile | 1.453 ms | 0 - 69 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Snapdragon® 7 Gen 4 Mobile | 3.844 ms | 0 - 189 MB | NPU
| FFNet-122NS-LowRes | TFLITE | w8a8 | Snapdragon® 8 Elite Gen 5 Mobile | 1.304 ms | 0 - 76 MB | NPU

## License
* The license for the original implementation of FFNet-122NS-LowRes can be found
  [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE).

## References
* [Simple and Efficient Architectures for Semantic Segmentation](https://arxiv.org/abs/2206.08236)
* [Source Model Implementation](https://github.com/Qualcomm-AI-research/FFNet)

## Community
* Join [our AI Hub Slack community](https://aihub.qualcomm.com/community/slack) to collaborate, post questions and learn more about on-device AI.
* For questions or feedback please [reach out to us](mailto:ai-hub-support@qti.qualcomm.com).