SSDLite MobileNet V2 — OpenVINO FP16
This repository distributes an existing OpenVINO IR conversion of the
TensorFlow SSDLite MobileNet V2 COCO object-detection model for the Advantech
Intel EdgeAI SDK benchmark. Advantech did not train the original model.
Files
| File |
Purpose |
ssdlite_mobilenet_v2.xml |
OpenVINO IR v11 graph and conversion metadata |
ssdlite_mobilenet_v2.bin |
Model constants, including FP16-compressed weights |
SHA256SUMS |
SHA256 checksums of the XML and BIN files |
LICENSE |
Upstream TensorFlow Models Apache-2.0 license |
Download both model files into the same directory and retain their filenames.
FP16 describes the stored weights; it does not guarantee that all operations
execute with FP16 precision on every device.
Model interface
| Tensor |
Type |
Shape |
Description |
Input image_tensor |
U8 |
[1, 300, 300, 3] |
NHWC image; this conversion expects BGR channel order |
Output detection_boxes |
F32 |
[1, 1, 100, 7] |
SSD detections: image ID, class ID, confidence, xmin, ymin, xmax, ymax |
The tensor names and shapes describe the XML in this repository. Other upstream
conversions can use different output names or preprocessing.
Benchmark use
The original artifact metadata records OpenVINO 2024.6.0 conversion tooling.
The existing XML and BIN were retained without reconversion and tested with
OpenVINO 2026.3.0 on CPU, Intel iGPU and Intel NPU in the Panther Lake Linux SDK.
With an OpenVINO C++ benchmark_app and its runtime available:
benchmark_app -m ssdlite_mobilenet_v2.xml -d CPU -hint throughput -t 10
benchmark_app -m ssdlite_mobilenet_v2.xml -d GPU -hint throughput -t 10
benchmark_app -m ssdlite_mobilenet_v2.xml -d NPU -hint throughput -t 10
Verify downloaded files before use:
sha256sum -c SHA256SUMS
Random-input benchmark throughput does not measure COCO detection accuracy or
end-to-end video FPS. No accuracy re-evaluation is claimed for this conversion.
Device support and performance depend on the installed runtime and drivers.
This repository provides files for local OpenVINO execution.
Sources and license
The upstream model license is Apache-2.0. The upstream license text is retained
in LICENSE; conversion metadata remains in the model XML.
