BGNet: Optimized for Mobile Deployment

Segment images in real-time on device

BGNet or Boundary-Guided Network, is a model designed for camouflaged object detection. It leverages edge semantics to enhance the representation learning process, making it more effective at identifying objects that blend into their surroundings

This model is an implementation of BGNet found here.

More details on model performance across various devices, can be found here.

Model Details

Model Type: Semantic segmentation
Model Stats:
- Model checkpoint: BGNet
- Input resolution: 416x416
- Number of parameters: 77.8M
- Model size: 297 MB

Model	Device	Chipset	Target Runtime	Inference Time (ms)	Peak Memory Range (MB)	Precision	Primary Compute Unit	Target Model
BGNet	Samsung Galaxy S23	Snapdragon® 8 Gen 2	TFLITE	22.861 ms	1 - 19 MB	FP16	NPU	--
BGNet	Samsung Galaxy S23	Snapdragon® 8 Gen 2	QNN	24.29 ms	2 - 18 MB	FP16	NPU	--
BGNet	Samsung Galaxy S23	Snapdragon® 8 Gen 2	ONNX	20.576 ms	0 - 334 MB	FP16	NPU	--
BGNet	Samsung Galaxy S24	Snapdragon® 8 Gen 3	TFLITE	16.997 ms	0 - 233 MB	FP16	NPU	--
BGNet	Samsung Galaxy S24	Snapdragon® 8 Gen 3	QNN	17.811 ms	218 - 293 MB	FP16	NPU	--
BGNet	Samsung Galaxy S24	Snapdragon® 8 Gen 3	ONNX	14.894 ms	2 - 79 MB	FP16	NPU	--
BGNet	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	TFLITE	15.453 ms	1 - 127 MB	FP16	NPU	--
BGNet	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	QNN	16.676 ms	2 - 69 MB	FP16	NPU	--
BGNet	Snapdragon 8 Elite QRD	Snapdragon® 8 Elite	ONNX	15.82 ms	3 - 66 MB	FP16	NPU	--
BGNet	SA7255P ADP	SA7255P	TFLITE	855.371 ms	1 - 126 MB	FP16	NPU	--
BGNet	SA7255P ADP	SA7255P	QNN	871.97 ms	2 - 12 MB	FP16	NPU	--
BGNet	SA8255 (Proxy)	SA8255P Proxy	TFLITE	22.922 ms	1 - 20 MB	FP16	NPU	--
BGNet	SA8255 (Proxy)	SA8255P Proxy	QNN	20.043 ms	2 - 4 MB	FP16	NPU	--
BGNet	SA8295P ADP	SA8295P	TFLITE	37.962 ms	1 - 99 MB	FP16	NPU	--
BGNet	SA8295P ADP	SA8295P	QNN	34.723 ms	2 - 19 MB	FP16	NPU	--
BGNet	SA8650 (Proxy)	SA8650P Proxy	TFLITE	22.977 ms	0 - 19 MB	FP16	NPU	--
BGNet	SA8650 (Proxy)	SA8650P Proxy	QNN	19.884 ms	0 - 2 MB	FP16	NPU	--
BGNet	SA8775P ADP	SA8775P	TFLITE	42.234 ms	1 - 126 MB	FP16	NPU	--
BGNet	SA8775P ADP	SA8775P	QNN	39.843 ms	2 - 12 MB	FP16	NPU	--
BGNet	QCS8275 (Proxy)	QCS8275 Proxy	TFLITE	855.371 ms	1 - 126 MB	FP16	NPU	--
BGNet	QCS8275 (Proxy)	QCS8275 Proxy	QNN	871.97 ms	2 - 12 MB	FP16	NPU	--
BGNet	QCS8550 (Proxy)	QCS8550 Proxy	TFLITE	22.842 ms	0 - 19 MB	FP16	NPU	--
BGNet	QCS8550 (Proxy)	QCS8550 Proxy	QNN	19.844 ms	2 - 4 MB	FP16	NPU	--
BGNet	QCS9075 (Proxy)	QCS9075 Proxy	TFLITE	42.234 ms	1 - 126 MB	FP16	NPU	--
BGNet	QCS9075 (Proxy)	QCS9075 Proxy	QNN	39.843 ms	2 - 12 MB	FP16	NPU	--
BGNet	QCS8450 (Proxy)	QCS8450 Proxy	TFLITE	33.789 ms	1 - 213 MB	FP16	NPU	--
BGNet	QCS8450 (Proxy)	QCS8450 Proxy	QNN	36.41 ms	0 - 49 MB	FP16	NPU	--
BGNet	Snapdragon X Elite CRD	Snapdragon® X Elite	QNN	20.351 ms	2 - 2 MB	FP16	NPU	--
BGNet	Snapdragon X Elite CRD	Snapdragon® X Elite	ONNX	22.249 ms	154 - 154 MB	FP16	NPU	--

License

The license for the original implementation of BGNet can be found [here](This model's original implementation does not provide a LICENSE.).
The license for the compiled assets for on-device deployment can be found here

References

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