Recherche avancée

Sur d’autres sites (3009)

• bwdif_vulkan : convert to storage images

  17 février, par Lynne

  bwdif_vulkan : convert to storage images
  texture() uses bilinear scaling ; imageLoad() accesses the image directly.
  
  The reason why texture() was used throughout Vulkan filters is that
  
  back when they were written, they were targetting old Intel hardware,
  
  which had a texel cache only for sampled images.
  These days, GPUs have a generic cache that doesn't care what source it
  
  gets populated with. Additionally, bypassing the sampling circuitry saves
  
  us some performance.
  Finally, all the old texture() code had an issue where unnormalized
  
  coordinates were used, but an offset of 0.5 was not added, hence each
  
  pixel ended up being interpolated. This fixes this.
  

  • [DH] libavfilter/vf_bwdif_vulkan.c
  • [DH] libavfilter/vulkan/bwdif.comp

• hwcontext_vulkan : enable read/write without storage

  22 février, par Lynne

  hwcontext_vulkan : enable read/write without storage
  BGR formats in Vulkan cannot be used in storage images, as the
  
  pixel labels on storage images are always ordered as RGB, and
  
  swizzling is not an option due to old hardware limitations.
  
  This means that you must always use an RGB format and manually
  
  swizzle when reading or writing to BGR images, or simply not use
  
  a format in the shader itself.
  
  This adds support for the latter.
  

  • [DH] libavutil/hwcontext_vulkan.c

• Connect a remote Ip camera as a Webrtc client

  5 avril 2017, par idosh

  I have 2 cameras :

  • An internal webcam embedded in my laptop.
  • A remote IP camera that is connected to my laptop through Wifi (transmits TCP, raw H264 data - no container). I’m getting the stream using node.js.

  My goal is to create a Webrtc network and connect the remote camera as another client.

  I’m trying to figure out possible solutions :

  • My naive thinking was that I would stream the remote camera payload to the browser. But as I came to understand the browser can’t handle the stream without a container. Fair enough. But I don’t understand why it does handle the video stream that arrives from my internal camera (from the navigator.getUserMedia() function). what’s the difference between the two streams ? why can’t I mimic the stream from the remote camera as the input ?
  • To bypass this problem I thought about creating a virtual camera using Manycam (or Manycam like app). To accomplish that I need to convert my TCP stream into an RTP stream (in order to feed Manycam). Though I did saw some info in ffmpeg command line, I couldn’t find info in their node.js api package "fluent-ffmpeg". Is it possible to do it using fluent-ffmpeg ? Or only using the command line tool ? Would it require another rtp server in the middle such as this one ?.
  • Third option I read about is using node.js as a client in Webrtc. I saw it was implemented in "simple-peer". I tried it out using their co-work with socket.io (socket.io-p2p). unfortunately I couldn’t get it to work / : When i’m trying to create a socket/peer in the server - it throws errors, as it expect options that are only available on the client-side (like window, location, etc.). Am I doing something wrong ? maybe there is more suitable framework for this matter ?
  • Forth option is to use a streaming server in the middle such as Kurnto. From my understanding it receives rtp as an input and transmits it as a webrtc client. I feel it’s the most excessive option, but maybe it’s not so bad (I have to admit that I haven’t investigate this option yet).

  any thoughts ?

  thanks !