capture frame with sdk



We need to programmatically capture 10 jpg frame/sec from GXV3651FHD in order to obtain a short flv videos using jpgs+ffmpeg converter.
Since HTTP API is to much slow for the purpose, I try to use SDK using GSNET_ClientCapturePicture API.
Program work, but only first 10-12 pictures is ok, subequent have “HISILICON” marker.
Is it possible to remove this?

We also try to save in H264 and MJPEG format (setting primary stream properly and using GSNET_ClientStartRecord) and try to convert with ffmpeg but the conversion fail. (result files format is non-standard H264/MJPEG for ffmpeg).



It’s a lot easier to do this with the HTTP API:

To capture an image:
Primary stream is view0.jpg
GXV3504 primary streams are view0.jpg – view3.jpg. For secondary streams use view4.jpg to view7.jpg

To pull the MJPEG stream:
Just replace the IP with yours.


That looks fantastic, I want to purchase a GXV3504 to be able to read 4 x MJPEG streams from custom written DVR software. What kind of framerate can I expect if I’m reading all 4 channels at D1 resolution from:

At the same time?


Was a solution ever found to this?
It is completely unacceptable that the writers would put a watermark on an image after customers paid for the hardware it’s accessing. I need this for image analysis, so the watermark is corrupting my data.
Recommending not using the SDK is not a solution either, although I will try the HTTP API to see if it is comparable.