OK, so, new problem.
In case anyone still doesn’t know, I’m an OpenVINO user. 
OpenVINO isn’t working. It’s not working in Sharpen AI 3.0.0, and it’s not working in Denoise AI 3.0.2 either.
I’m on Windows 10 Pro x64 2004.
I ran an image on Sharpen AI version 2.2.1 on Windows 7. It took about 5 - 10 seconds or so.
The same image on Sharpen AI version 3.0.2 on Windows 10 took over 40 seconds.
I have the latest beta of Video Enhance AI installed on the same system, and there, OpenVINO works extremely well.
Attached below are the logs from Windows 10.
2021-03-25-01-28-34-Main.tzlog (67.5 KB)
2021-03-25-01-28-34-QML.tzlog (25.8 KB)
2021-03-25-01-28-34-Main.tzlog seems to show it found the system can support OpenVINO. But check lines 391 - 412, it says:
2021-03-25T01:30:05.342 [Info ] Thread: 0x1edb485ff20 < line #: 0> CAN DOWNLOAD MODELS: true
2021-03-25T01:30:05.342 [Info ] Thread: 0x1edb485ff20 < line #: 0> BACKENDS: coreml, onnx16, openvino16, tensorflow
2021-03-25T01:30:05.342 [Debug ] Thread: 0x1edb485ff20 < line #: 0> CHOOSE FROM BACKENDS: 146
2021-03-25T01:30:05.342 [Info ] Thread: 0x1edb485ff20 < line #: 0> Device ID: -1 FOUND OV: false CML: false TF: true ONNX: false ONNX16: true
2021-03-25T01:30:05.342 [Info ] Thread: 0x1edb485ff20 < line #: 0> GPU NAME:
2021-03-25T01:30:05.342 [Debug ] Thread: 0x1edb485ff20 < line #: 0> CPU MODE: 1
2021-03-25T01:30:05.342 [Debug ] Thread: 0x1edb485ff20 < line #: 0> Block: 512 512 12 1.16038 0.560381
2021-03-25T01:30:05.342 [Info ] Thread: 0x1edb485ff20 < line #: 0> BLOCK: 512 512
2021-03-25T01:30:05.342 [Debug ] Thread: 0x1edb485ff20 < line #: 0> [N]-v[V]-fp16-[H]x[W]-ox.tz FILENAME: slb-v1-fp16-512x512-ox.tz
2021-03-25T01:30:05.342 [Info ] Thread: 0x1edb485ff20 < line #: 0> DEVICE: -1
2021-03-25T01:30:05.342 [Debug ] Thread: 0x1edb485ff20 < line #: 0> Loading default model file: C:\Users\Admin\AppData\Local\Topaz Labs\Topaz Sharpen AI\tgrc\/slb-v1-fp16-512x512-ox.tz
2021-03-25T01:30:05.342 [Debug ] Thread: 0x1edb485ff20 < line #: 0> Trying to open file: C:\Users\Admin\AppData\Local\Topaz Labs\Topaz Sharpen AI\tgrc\/slb-v1-fp16-512x512-ox.tz
2021-03-25T01:30:05.342 [Debug ] Thread: 0x1edb485ff20 < line #: 0> Opened archive containing 1 files
2021-03-25T01:30:06.659 [Info ] Thread: 0x1edb485ff20 < line #: 0> OX with device: -1
2021-03-25T01:30:06.659 [Info ] Thread: 0x1edb485ff20 < line #: 0> BUFFER DATA: 75732786
2021-03-25T01:30:06.659 [Info ] Thread: 0x1edb485ff20 < line #: 0> ORT ENV SETUP
2021-03-25T01:30:06.755 [Info ] Thread: 0x1edb485ff20 < line #: 0> Loading time for model file: C:\Users\Admin\AppData\Local\Topaz Labs\Topaz Sharpen AI\tgrc\/slb-v1-fp16-512x512-ox.tz is 1413267
2021-03-25T01:30:06.759 [Info ] Thread: 0x1edb485ff20 < line #: 0> TOTAL INSTANCES: 1
2021-03-25T01:30:06.759 [Info ] Thread: 0x1edb485ff20 < line #: 0> Model : 0x1edc79d37e0
2021-03-25T01:30:06.759 [Info ] Thread: 0x1edb485ff20 < line #: 0> Block Size selected: 512 x 512
2021-03-25T01:30:06.759 [Debug ] Thread: 0x1edb485ff20 < line #: 0> Clamping required false
2021-03-25T01:30:06.759 [Debug ] Thread: 0x1edb485ff20 < line #: 0> MT created
Line 394 says “FOUND OV: false” and you can see it’s not loading the ov.tz files; it’s loading the tensorflow ox.tz files.
The application is set to “CPU” for “AI Processor” in Preferences.
In the registry, under HKEY_CURRENT_USER\SOFTWARE\Topaz Labs LLC\Topaz Sharpen AI\appMain, DWORD cpuPlugin is set to 1. I believe it should be 0 for OpenVINO. If I change it to 0 it will just get changed back to 1. However, String ifOpenVinoCanWork is set to true.
For comparison, here are logs for Sharpen AI 2.2.1 on Windows Server 2008 R2 (effectively Windows 7), where OpenVINO is working:
2021-03-25-01-55-47-Main.tzlog (44.2 KB)
2021-03-25-01-55-47-QML.tzlog (24.6 KB)
Video Enhance AI version 1.5.2 had a similar problem, where it used Tensorflow instead of OpenVINO, even though the system could support OpenVINO. This was fixed in Video Enhance AI version 1.6.0. I had a support ticket for this, it’s ticket # 250486. However, I don’t think this will help much, as Matt Lathrop, the developer who fixed it, is no longer with Topaz Labs. And it may not have the same cause, even though the issue seems similar. But I thought I’d mention it.
