Yolo-NAS-Quantized: Optimized for Mobile Deployment
Quantized real-time object detection optimized for mobile and edge
YoloNAS is a machine learning model that predicts bounding boxes and classes of objects in an image. This model is post-training quantized to int8 using samples from the COCO dataset.
This model is an implementation of Yolo-NAS-Quantized found here.
More details on model performance accross various devices, can be found here.
Model Details
- Model Type: Object detection
- Model Stats:
- Model checkpoint: YoloNAS Small
- Input resolution: 640x640
- Number of parameters: 12.2M
- Model size: 12.1 MB
| Model | Device | Chipset | Target Runtime | Inference Time (ms) | Peak Memory Range (MB) | Precision | Primary Compute Unit | Target Model |
|---|---|---|---|---|---|---|---|---|
| Yolo-NAS-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | TFLITE | 4.693 ms | 0 - 23 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | Samsung Galaxy S23 | Snapdragon® 8 Gen 2 | QNN | 4.307 ms | 0 - 13 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | TFLITE | 3.057 ms | 0 - 31 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | Samsung Galaxy S24 | Snapdragon® 8 Gen 3 | QNN | 2.915 ms | 1 - 32 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | TFLITE | 3.133 ms | 0 - 29 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | Snapdragon 8 Elite QRD | Snapdragon® 8 Elite | QNN | 2.447 ms | 0 - 28 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | TFLITE | 14.75 ms | 0 - 37 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | RB3 Gen 2 (Proxy) | QCS6490 Proxy | QNN | 15.376 ms | 1 - 8 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | TFLITE | 4.683 ms | 0 - 20 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | QCS8550 (Proxy) | QCS8550 Proxy | QNN | 4.055 ms | 1 - 3 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA7255P ADP | SA7255P | TFLITE | 33.757 ms | 0 - 26 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA7255P ADP | SA7255P | QNN | 33.028 ms | 1 - 11 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA8255 (Proxy) | SA8255P Proxy | TFLITE | 4.788 ms | 0 - 17 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA8255 (Proxy) | SA8255P Proxy | QNN | 4.063 ms | 3 - 4 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA8295P ADP | SA8295P | TFLITE | 6.586 ms | 0 - 24 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA8295P ADP | SA8295P | QNN | 6.03 ms | 1 - 7 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA8650 (Proxy) | SA8650P Proxy | TFLITE | 4.781 ms | 0 - 19 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA8650 (Proxy) | SA8650P Proxy | QNN | 4.062 ms | 4 - 5 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA8775P ADP | SA8775P | TFLITE | 6.472 ms | 0 - 29 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | SA8775P ADP | SA8775P | QNN | 5.594 ms | 1 - 7 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | TFLITE | 5.248 ms | 0 - 37 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | QCS8450 (Proxy) | QCS8450 Proxy | QNN | 4.693 ms | 1 - 35 MB | INT8 | NPU | -- |
| Yolo-NAS-Quantized | Snapdragon X Elite CRD | Snapdragon® X Elite | QNN | 4.443 ms | 1 - 1 MB | INT8 | NPU | -- |
License
- The license for the original implementation of Yolo-NAS-Quantized can be found here.
- The license for the compiled assets for on-device deployment can be found here
References
- YOLO-NAS by Deci Achieves SOTA Performance on Object Detection Using Neural Architecture Search
- Source Model Implementation
Community
- Join our AI Hub Slack community to collaborate, post questions and learn more about on-device AI.
- For questions or feedback please reach out to us.
Usage and Limitations
Model may not be used for or in connection with any of the following applications:
- Accessing essential private and public services and benefits;
- Administration of justice and democratic processes;
- Assessing or recognizing the emotional state of a person;
- Biometric and biometrics-based systems, including categorization of persons based on sensitive characteristics;
- Education and vocational training;
- Employment and workers management;
- Exploitation of the vulnerabilities of persons resulting in harmful behavior;
- General purpose social scoring;
- Law enforcement;
- Management and operation of critical infrastructure;
- Migration, asylum and border control management;
- Predictive policing;
- Real-time remote biometric identification in public spaces;
- Recommender systems of social media platforms;
- Scraping of facial images (from the internet or otherwise); and/or
- Subliminal manipulation
Inference API (serverless) does not yet support pytorch models for this pipeline type.