The package contains the test program and pre-compiled dynamic library. It only includes the Windows version.

When running the test program for the first time, the window size may be small. Clicking the close button will restore it to normal size. To exit the program, press the 'q' or 'esc' key.

Update (v1.1): pressing the 'A' key now triggers automatic segmentation (takes approximately one minute to process a single image).

Update (v1.2): allows for the addition of foreground points by holding down the Control key and left-clicking, right-clicking enables the addition of background points.

Update (v1.3): CUDA support is now included. when running the example program from the command line, add parameter -sam_device="cuda:0" to use the first graphics card.

Update (v1.4): box input is supported for segmentation (Issue #6). Middle mouse drag creates a box.
Update (v1.4.1): Adding MobileSAM as a backend model (.\sam_cpp_test.exe -pre_model="models/mobile_sam_preprocess.onnx" -sam_model="models/mobile_sam.onnx"). If only MobileSAM is needed, you can remove the "sam_preprocess.onnx" and "sam_vit_h_4b8939.onnx" files from the Models folder to significantly reduce the overall size.
v1.5: Support HQ-SAM. Command line usage: [HQ-SAM] sam_cpp_test.exe -pre_model="models/sam_hq_preprocess.onnx" -sam_model="models/sam_hq_vit_h.onnx" -image="images/example4.png", [SAM] sam_cpp_test.exe -pre_model="models/sam_preprocess.onnx" -sam_model="models/sam_vit_h_4b8939.onnx" -image="images/example4.png"

v1.6: support EdgeSAM. sam_cpp_test -pre_model="models/edge_sam_3x_encoder.onnx" -sam_model="models/edge_sam_3x_decoder.onnx". ONNX files in the package are downloaded from the official page. Currently, the input image size is set to 1024x1024.
If support for other sizes is needed, encoder.onnx need to be re-export using the export_onnx_model.py.

The package contains the test program and pre-compiled dynamic library. It only includes the Windows version.

When running the test program for the first time, the window size may be small. Clicking the close button will restore it to normal size. To exit the program, press the 'q' or 'esc' key.

Update (v1.1): pressing the 'A' key now triggers automatic segmentation (takes approximately one minute to process a single image).

Update (v1.2): allows for the addition of foreground points by holding down the Control key and left-clicking, right-clicking enables the addition of background points.

Update (v1.3): CUDA support is now included. when running the example program from the command line, add parameter -sam_device="cuda:0" to use the first graphics card.

Update (v1.4): box input is supported for segmentation (Issue #6). Middle mouse drag creates a box.

Update (v1.4.1): Adding MobileSAM as a backend model (.\sam_cpp_test.exe -pre_model="models/mobile_sam_preprocess.onnx" -sam_model="models/mobile_sam.onnx"). If only MobileSAM is needed, you can remove the "sam_preprocess.onnx" and "sam_vit_h_4b8939.onnx" files from the Models folder to significantly reduce the overall size.

v1.5: Support HQ-SAM. Command line usage: [HQ-SAM] sam_cpp_test.exe -pre_model="models/sam_hq_preprocess.onnx" -sam_model="models/sam_hq_vit_h.onnx" -image="images/example4.png", [SAM] sam_cpp_test.exe -pre_model="models/sam_preprocess.onnx" -sam_model="models/sam_vit_h_4b8939.onnx" -image="images/example4.png"

The package contains the test program and pre-compiled dynamic library. It only includes the Windows version.

When running the test program for the first time, the window size may be small. Clicking the close button will restore it to normal size. To exit the program, press the 'q' or 'esc' key.

Update (v1.1): pressing the 'A' key now triggers automatic segmentation (takes approximately one minute to process a single image).

Update (v1.2): allows for the addition of foreground points by holding down the Control key and left-clicking, right-clicking enables the addition of background points.

Update (v1.3): CUDA support is now included. when running the example program from the command line, add parameter -sam_device="cuda:0" to use the first graphics card.

Update (v1.4): box input is supported for segmentation (Issue #6). Middle mouse drag creates a box.

Update (v1.4.1): Adding MobileSAM as a backend model (.\sam_cpp_test.exe -pre_model="models/mobile_sam_preprocess.onnx" -sam_model="models/mobile_sam.onnx"). If only MobileSAM is needed, you can remove the "sam_preprocess.onnx" and "sam_vit_h_4b8939.onnx" files from the Models folder to significantly reduce the overall size.

The package contains the test program and pre-compiled dynamic library. It only includes the Windows version.

When running the test program for the first time, the window size may be small. Clicking the close button will restore it to normal size. To exit the program, press the 'q' or 'esc' key.

Update (v1.1): pressing the 'A' key now triggers automatic segmentation (takes approximately one minute to process a single image).

Update (v1.2): allows for the addition of foreground points by holding down the Control key and left-clicking, right-clicking enables the addition of background points.

Update (v1.3): CUDA support is now included. when running the example program from the command line, add parameter -sam_device="cuda:0" to use the first graphics card.

Update (v1.4): box input is supported for segmentation (Issue #6). Middle mouse drag creates a box.

The package contains the test program and pre-compiled dynamic library. It only includes the Windows version.

When running the test program for the first time, the window size may be small. Clicking the close button will restore it to normal size. To exit the program, press the 'q' or 'esc' key.

Update (v1.1): pressing the 'A' key now triggers automatic segmentation (takes approximately one minute to process a single image).

Update (v1.2): allows for the addition of foreground points by holding down the Control key and left-clicking, right-clicking enables the addition of background points.

Update (v1.3): CUDA support is now included. when running the example program from the command line, add parameter -sam_device="cuda:0" to use the first graphics card.

The package contains the test program and pre-compiled dynamic library. It only includes the Windows version.

When running the test program for the first time, the window size may be small. Clicking the close button will restore it to normal size. To exit the program, press the 'q' or 'esc' key.

Update (v1.1): pressing the 'A' key now triggers automatic segmentation (takes approximately one minute to process a single image).

Update since previous relese: allows for the addition of foreground points by holding down the Control key and left-clicking, right-clicking enables the addition of background points.

The package contains the test program and pre-compiled dynamic library. It only includes the Windows version.

When running the test program for the first time, the window size may be small. Clicking the close button will restore it to normal size. To exit the program, press the 'q' or 'esc' key.

Update since previous relese: pressing the 'A' key now triggers automatic segmentation (takes approximately one minute to process a single image).

The package contains the test program and pre-compiled dynamic library. It only includes the Windows version.

When running the test program for the first time, the window size may be small. Clicking the close button will restore it to normal size. To exit the program, press the 'q' or 'esc' key.