Hugging Face's logo

qualcomm
/
YOLOv8-Segmentation

Image Segmentation
PyTorch
real_time
android
Model card Files
xet
 Community
1
qaihm-bot commited on
Commit
ac9a230
·
verified ·
1 Parent(s): 9e63900

v0.48.0

Browse files

See https://github.com/qualcomm/ai-hub-models/releases/v0.48.0 for changelog.

Files changed (1) hide show
  1. README.md +32 -32
README.md CHANGED
@@ -15,18 +15,18 @@ pipeline_tag: image-segmentation
15
  Ultralytics YOLOv8 is a machine learning model that predicts bounding boxes, segmentation masks and classes of objects in an image.
16
 
17
  This is based on the implementation of YOLOv8-Segmentation found [here](https://github.com/ultralytics/ultralytics/tree/main/ultralytics/models/yolo/segment).
18
- 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/yolov8_seg) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
19
 
20
  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.
21
 
22
  ## Getting Started
23
  Due to licensing restrictions, we cannot distribute pre-exported model assets for this model.
24
- Use the [Qualcomm® AI Hub Models](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov8_seg) Python library to compile and export the model with your own:
25
  - Custom weights (e.g., fine-tuned checkpoints)
26
  - Custom input shapes
27
  - Target device and runtime configurations
28
 
29
- See our repository for [YOLOv8-Segmentation on GitHub](https://github.com/quic/ai-hub-models/blob/main/qai_hub_models/models/yolov8_seg) for usage instructions.
30
 
31
 
32
  ## Model Details
@@ -44,35 +44,35 @@ See our repository for [YOLOv8-Segmentation on GitHub](https://github.com/quic/a
44
  ## Performance Summary
45
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
46
  |---|---|---|---|---|---|---
47
- | YOLOv8-Segmentation | ONNX | float | Snapdragon® X Elite | 6.899 ms | 17 - 17 MB | NPU
48
- | YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 4.058 ms | 17 - 293 MB | NPU
49
- | YOLOv8-Segmentation | ONNX | float | Qualcomm® QCS8550 (Proxy) | 6.338 ms | 12 - 20 MB | NPU
50
- | YOLOv8-Segmentation | ONNX | float | Qualcomm® QCS9075 | 7.827 ms | 12 - 15 MB | NPU
51
- | YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 3.313 ms | 0 - 221 MB | NPU
52
- | YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.957 ms | 0 - 233 MB | NPU
53
- | YOLOv8-Segmentation | ONNX | float | Snapdragon® X2 Elite | 3.429 ms | 16 - 16 MB | NPU
54
- | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® X Elite | 4.824 ms | 5 - 5 MB | NPU
55
- | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 3.34 ms | 0 - 213 MB | NPU
56
- | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 16.945 ms | 0 - 179 MB | NPU
57
- | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 4.479 ms | 5 - 117 MB | NPU
58
- | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA8775P | 6.288 ms | 0 - 183 MB | NPU
59
- | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS9075 | 6.03 ms | 5 - 15 MB | NPU
60
- | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 9.89 ms | 5 - 196 MB | NPU
61
- | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA7255P | 16.945 ms | 0 - 179 MB | NPU
62
- | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA8295P | 9.224 ms | 0 - 166 MB | NPU
63
- | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.654 ms | 0 - 185 MB | NPU
64
- | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.908 ms | 4 - 189 MB | NPU
65
- | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® X2 Elite | 2.798 ms | 5 - 5 MB | NPU
66
- | YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.951 ms | 0 - 110 MB | NPU
67
- | YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 16.139 ms | 4 - 84 MB | NPU
68
- | YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 3.929 ms | 0 - 13 MB | NPU
69
- | YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA8775P | 5.856 ms | 4 - 89 MB | NPU
70
- | YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS9075 | 5.809 ms | 4 - 23 MB | NPU
71
- | YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 8.926 ms | 0 - 200 MB | NPU
72
- | YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA7255P | 16.139 ms | 4 - 84 MB | NPU
73
- | YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA8295P | 8.651 ms | 4 - 174 MB | NPU
74
- | YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.247 ms | 0 - 80 MB | NPU
75
- | YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.766 ms | 0 - 102 MB | NPU
76
 
77
  ## License
78
  * The license for the original implementation of YOLOv8-Segmentation can be found
 
15
  Ultralytics YOLOv8 is a machine learning model that predicts bounding boxes, segmentation masks and classes of objects in an image.
16
 
17
  This is based on the implementation of YOLOv8-Segmentation found [here](https://github.com/ultralytics/ultralytics/tree/main/ultralytics/models/yolo/segment).
18
+ This repository contains pre-exported model files optimized for Qualcomm® devices. You can use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/yolov8_seg) library to export with custom configurations. More details on model performance across various devices, can be found [here](#performance-summary).
19
 
20
  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.
21
 
22
  ## Getting Started
23
  Due to licensing restrictions, we cannot distribute pre-exported model assets for this model.
24
+ Use the [Qualcomm® AI Hub Models](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/yolov8_seg) Python library to compile and export the model with your own:
25
  - Custom weights (e.g., fine-tuned checkpoints)
26
  - Custom input shapes
27
  - Target device and runtime configurations
28
 
29
+ See our repository for [YOLOv8-Segmentation on GitHub](https://github.com/qualcomm/ai-hub-models/blob/main/qai_hub_models/models/yolov8_seg) for usage instructions.
30
 
31
 
32
  ## Model Details
 
44
  ## Performance Summary
45
  | Model | Runtime | Precision | Chipset | Inference Time (ms) | Peak Memory Range (MB) | Primary Compute Unit
46
  |---|---|---|---|---|---|---
47
+ | YOLOv8-Segmentation | ONNX | float | Snapdragon® X2 Elite | 3.422 ms | 16 - 16 MB | NPU
48
+ | YOLOv8-Segmentation | ONNX | float | Snapdragon® X Elite | 6.888 ms | 17 - 17 MB | NPU
49
+ | YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Gen 3 Mobile | 4.045 ms | 14 - 288 MB | NPU
50
+ | YOLOv8-Segmentation | ONNX | float | Qualcomm® QCS8550 (Proxy) | 6.355 ms | 12 - 19 MB | NPU
51
+ | YOLOv8-Segmentation | ONNX | float | Qualcomm® QCS9075 | 7.778 ms | 15 - 18 MB | NPU
52
+ | YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Elite For Galaxy Mobile | 3.317 ms | 2 - 227 MB | NPU
53
+ | YOLOv8-Segmentation | ONNX | float | Snapdragon® 8 Elite Gen 5 Mobile | 2.956 ms | 0 - 233 MB | NPU
54
+ | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® X2 Elite | 2.737 ms | 5 - 5 MB | NPU
55
+ | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® X Elite | 4.84 ms | 5 - 5 MB | NPU
56
+ | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Gen 3 Mobile | 3.331 ms | 5 - 216 MB | NPU
57
+ | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS8275 (Proxy) | 16.901 ms | 1 - 180 MB | NPU
58
+ | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS8550 (Proxy) | 4.483 ms | 5 - 6 MB | NPU
59
+ | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA8775P | 6.307 ms | 1 - 183 MB | NPU
60
+ | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS9075 | 6.057 ms | 5 - 15 MB | NPU
61
+ | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® QCS8450 (Proxy) | 9.893 ms | 5 - 196 MB | NPU
62
+ | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA7255P | 16.901 ms | 1 - 180 MB | NPU
63
+ | YOLOv8-Segmentation | QNN_DLC | float | Qualcomm® SA8295P | 9.217 ms | 0 - 166 MB | NPU
64
+ | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.65 ms | 5 - 190 MB | NPU
65
+ | YOLOv8-Segmentation | QNN_DLC | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.912 ms | 4 - 190 MB | NPU
66
+ | YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Gen 3 Mobile | 2.939 ms | 0 - 114 MB | NPU
67
+ | YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS8275 (Proxy) | 16.141 ms | 4 - 84 MB | NPU
68
+ | YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS8550 (Proxy) | 3.946 ms | 4 - 9 MB | NPU
69
+ | YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA8775P | 5.846 ms | 4 - 90 MB | NPU
70
+ | YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS9075 | 5.771 ms | 4 - 23 MB | NPU
71
+ | YOLOv8-Segmentation | TFLITE | float | Qualcomm® QCS8450 (Proxy) | 8.894 ms | 4 - 208 MB | NPU
72
+ | YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA7255P | 16.141 ms | 4 - 84 MB | NPU
73
+ | YOLOv8-Segmentation | TFLITE | float | Qualcomm® SA8295P | 8.606 ms | 4 - 174 MB | NPU
74
+ | YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Elite For Galaxy Mobile | 2.24 ms | 0 - 89 MB | NPU
75
+ | YOLOv8-Segmentation | TFLITE | float | Snapdragon® 8 Elite Gen 5 Mobile | 1.767 ms | 0 - 102 MB | NPU
76
 
77
  ## License
78
  * The license for the original implementation of YOLOv8-Segmentation can be found