Hugging Face's logo

qualcomm
/
FFNet-78S-Quantized

Image Segmentation
PyTorch
TF Lite
ONNX
quantized
real_time
android
Model card Files Community
qaihm-bot commited on
Commit
f3fa28e
·
verified ·
1 Parent(s): e64c6b0

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +35 -35
README.md CHANGED
@@ -37,48 +37,47 @@ More details on model performance across various devices, can be found
37
 
38
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
39
  |---|---|---|---|---|---|---|---|---|
40
- | FFNet-78S-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 11.938 ms | 1 - 18 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
41
- | FFNet-78S-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 23.76 ms | 6 - 19 MB | INT8 | NPU | [FFNet-78S-Quantized.so](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.so) |
42
- | FFNet-78S-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 15.3 ms | 0 - 59 MB | INT8 | NPU | [FFNet-78S-Quantized.onnx](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.onnx) |
43
- | FFNet-78S-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 8.673 ms | 1 - 38 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
44
- | FFNet-78S-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 17.191 ms | 6 - 42 MB | INT8 | NPU | [FFNet-78S-Quantized.so](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.so) |
45
- | FFNet-78S-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 11.011 ms | 8 - 67 MB | INT8 | NPU | [FFNet-78S-Quantized.onnx](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.onnx) |
46
- | FFNet-78S-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 7.363 ms | 0 - 37 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
47
- | FFNet-78S-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 18.192 ms | 6 - 42 MB | INT8 | NPU | Use Export Script |
48
- | FFNet-78S-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 10.474 ms | 6 - 55 MB | INT8 | NPU | [FFNet-78S-Quantized.onnx](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.onnx) |
49
- | FFNet-78S-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 51.11 ms | 1 - 41 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
50
- | FFNet-78S-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 93.889 ms | 6 - 17 MB | INT8 | NPU | Use Export Script |
51
- | FFNet-78S-Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 338.085 ms | 1 - 8 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
52
- | FFNet-78S-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 11.921 ms | 1 - 21 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
53
- | FFNet-78S-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 20.566 ms | 6 - 9 MB | INT8 | NPU | Use Export Script |
54
- | FFNet-78S-Quantized | SA7255P ADP | SA7255P | TFLITE | 157.142 ms | 1 - 31 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
55
- | FFNet-78S-Quantized | SA7255P ADP | SA7255P | QNN | 170.874 ms | 6 - 16 MB | INT8 | NPU | Use Export Script |
56
- | FFNet-78S-Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 11.922 ms | 1 - 15 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
57
- | FFNet-78S-Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 20.617 ms | 6 - 9 MB | INT8 | NPU | Use Export Script |
58
- | FFNet-78S-Quantized | SA8295P ADP | SA8295P | TFLITE | 19.095 ms | 0 - 36 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
59
- | FFNet-78S-Quantized | SA8295P ADP | SA8295P | QNN | 28.637 ms | 6 - 20 MB | INT8 | NPU | Use Export Script |
60
- | FFNet-78S-Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 12.086 ms | 0 - 23 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
61
- | FFNet-78S-Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 20.493 ms | 6 - 9 MB | INT8 | NPU | Use Export Script |
62
- | FFNet-78S-Quantized | SA8775P ADP | SA8775P | TFLITE | 15.254 ms | 1 - 31 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
63
- | FFNet-78S-Quantized | SA8775P ADP | SA8775P | QNN | 25.917 ms | 6 - 16 MB | INT8 | NPU | Use Export Script |
64
- | FFNet-78S-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 14.37 ms | 0 - 36 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
65
- | FFNet-78S-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 25.49 ms | 6 - 45 MB | INT8 | NPU | Use Export Script |
66
- | FFNet-78S-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 22.059 ms | 6 - 6 MB | INT8 | NPU | Use Export Script |
67
- | FFNet-78S-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 15.806 ms | 22 - 22 MB | INT8 | NPU | [FFNet-78S-Quantized.onnx](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.onnx) |
68
 
69
 
70
 
71
 
72
  ## Installation
73
 
74
- This model can be installed as a Python package via pip.
75
 
 
76
  ```bash
77
- pip install "qai-hub-models[ffnet_78s_quantized]"
78
  ```
79
 
80
 
81
-
82
  ## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
83
 
84
  Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
@@ -129,8 +128,8 @@ Profiling Results
129
  FFNet-78S-Quantized
130
  Device : Samsung Galaxy S23 (13)
131
  Runtime : TFLITE
132
- Estimated inference time (ms) : 11.9
133
- Estimated peak memory usage (MB): [1, 18]
134
  Total # Ops : 153
135
  Compute Unit(s) : NPU (153 ops)
136
  ```
@@ -157,7 +156,7 @@ from qai_hub_models.models.ffnet_78s_quantized import Model
157
  torch_model = Model.from_pretrained()
158
 
159
  # Device
160
- device = hub.Device("Samsung Galaxy S23")
161
 
162
  # Trace model
163
  input_shape = torch_model.get_input_spec()
@@ -235,7 +234,8 @@ Explore all available models on [Qualcomm® AI Hub](https://aihub.qualcomm.com/)
235
 
236
 
237
  ## License
238
- * The license for the original implementation of FFNet-78S-Quantized can be found [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE).
 
239
  * The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
240
 
241
 
 
37
 
38
  | Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model
39
  |---|---|---|---|---|---|---|---|---|
40
+ | FFNet-78S-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 12.056 ms | 1 - 17 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
41
+ | FFNet-78S-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 23.841 ms | 6 - 20 MB | INT8 | NPU | [FFNet-78S-Quantized.so](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.so) |
42
+ | FFNet-78S-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | ONNX | 15.52 ms | 1 - 55 MB | INT8 | NPU | [FFNet-78S-Quantized.onnx](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.onnx) |
43
+ | FFNet-78S-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 8.688 ms | 0 - 39 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
44
+ | FFNet-78S-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 17.248 ms | 6 - 41 MB | INT8 | NPU | [FFNet-78S-Quantized.so](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.so) |
45
+ | FFNet-78S-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | ONNX | 11.051 ms | 7 - 62 MB | INT8 | NPU | [FFNet-78S-Quantized.onnx](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.onnx) |
46
+ | FFNet-78S-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 8.815 ms | 1 - 39 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
47
+ | FFNet-78S-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 18.055 ms | 6 - 44 MB | INT8 | NPU | Use Export Script |
48
+ | FFNet-78S-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | ONNX | 9.988 ms | 7 - 57 MB | INT8 | NPU | [FFNet-78S-Quantized.onnx](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.onnx) |
49
+ | FFNet-78S-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 51.048 ms | 1 - 41 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
50
+ | FFNet-78S-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 94.03 ms | 6 - 17 MB | INT8 | NPU | Use Export Script |
51
+ | FFNet-78S-Quantized | RB5 (Proxy) | QCS8250 Proxy | TFLITE | 341.347 ms | 1 - 3 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
52
+ | FFNet-78S-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 11.926 ms | 1 - 21 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
53
+ | FFNet-78S-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 20.602 ms | 6 - 9 MB | INT8 | NPU | Use Export Script |
54
+ | FFNet-78S-Quantized | SA7255P ADP | SA7255P | TFLITE | 157.061 ms | 1 - 31 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
55
+ | FFNet-78S-Quantized | SA7255P ADP | SA7255P | QNN | 170.775 ms | 6 - 14 MB | INT8 | NPU | Use Export Script |
56
+ | FFNet-78S-Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 11.933 ms | 1 - 16 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
57
+ | FFNet-78S-Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 20.612 ms | 6 - 9 MB | INT8 | NPU | Use Export Script |
58
+ | FFNet-78S-Quantized | SA8295P ADP | SA8295P | TFLITE | 19.092 ms | 0 - 36 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
59
+ | FFNet-78S-Quantized | SA8295P ADP | SA8295P | QNN | 28.477 ms | 6 - 20 MB | INT8 | NPU | Use Export Script |
60
+ | FFNet-78S-Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 12.108 ms | 1 - 15 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
61
+ | FFNet-78S-Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 20.607 ms | 6 - 9 MB | INT8 | NPU | Use Export Script |
62
+ | FFNet-78S-Quantized | SA8775P ADP | SA8775P | TFLITE | 15.238 ms | 1 - 31 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
63
+ | FFNet-78S-Quantized | SA8775P ADP | SA8775P | QNN | 25.986 ms | 6 - 16 MB | INT8 | NPU | Use Export Script |
64
+ | FFNet-78S-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 14.358 ms | 1 - 37 MB | INT8 | NPU | [FFNet-78S-Quantized.tflite](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.tflite) |
65
+ | FFNet-78S-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 25.538 ms | 6 - 38 MB | INT8 | NPU | Use Export Script |
66
+ | FFNet-78S-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 21.572 ms | 6 - 6 MB | INT8 | NPU | Use Export Script |
67
+ | FFNet-78S-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | ONNX | 15.931 ms | 24 - 24 MB | INT8 | NPU | [FFNet-78S-Quantized.onnx](https://huggingface.co/qualcomm/FFNet-78S-Quantized/blob/main/FFNet-78S-Quantized.onnx) |
68
 
69
 
70
 
71
 
72
  ## Installation
73
 
 
74
 
75
+ Install the package via pip:
76
  ```bash
77
+ pip install "qai-hub-models[ffnet-78s-quantized]"
78
  ```
79
 
80
 
 
81
  ## Configure Qualcomm® AI Hub to run this model on a cloud-hosted device
82
 
83
  Sign-in to [Qualcomm® AI Hub](https://app.aihub.qualcomm.com/) with your
 
128
  FFNet-78S-Quantized
129
  Device : Samsung Galaxy S23 (13)
130
  Runtime : TFLITE
131
+ Estimated inference time (ms) : 12.1
132
+ Estimated peak memory usage (MB): [1, 17]
133
  Total # Ops : 153
134
  Compute Unit(s) : NPU (153 ops)
135
  ```
 
156
  torch_model = Model.from_pretrained()
157
 
158
  # Device
159
+ device = hub.Device("Samsung Galaxy S24")
160
 
161
  # Trace model
162
  input_shape = torch_model.get_input_spec()
 
234
 
235
 
236
  ## License
237
+ * The license for the original implementation of FFNet-78S-Quantized can be found
238
+ [here](https://github.com/Qualcomm-AI-research/FFNet/blob/master/LICENSE).
239
  * The license for the compiled assets for on-device deployment can be found [here](https://qaihub-public-assets.s3.us-west-2.amazonaws.com/qai-hub-models/Qualcomm+AI+Hub+Proprietary+License.pdf)
240
 
241