Recherche avancée

Sur d’autres sites (11608)

• avcodec/mpegvideo_enc : Reduce stack usage

  18 mai, par Andreas Rheinhardt

  avcodec/mpegvideo_enc : Reduce stack usage
  Multiple PutBitContexts are used when encoding partitioned
  
  frames. When there are multiple candidates for macroblock types,
  
  multiple states (namely the state before encoding the current MB,
  
  the best state among the ones already tried and the current one)
  
  need to be kept ; duplicates of the PutBitContexts are among this
  
  state. The temporary buffers for them are kept on the stack
  
  in encode_thread() and their size is quite generous (MAX_MB_SIZE
  
   3000 bytes). This commit uses tighter bounds, bringing the
  
  size of the pb2 buffer down to 15 bytes.
  Signed-off-by : Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
  

  • [DH] libavcodec/mpeg4videoenc.h
  • [DH] libavcodec/mpegvideo_enc.c

• Unable to retrieve video stream from RTSP URL inside Docker container

  6 février, par birdalugur

  I have a FastAPI application running inside a Docker container that is trying to stream video from an RTSP camera URL using OpenCV. The setup works fine locally, but when running inside Docker, the /video endpoint does not return a stream and times out. Below are the details of the issue.

  


  Docker Setup :

  


  Dockerfile :

  


  FROM python:3.10.12

RUN apt-get update && apt-get install -y \
    libgl1-mesa-glx \
    libglib2.0-0

WORKDIR /app

COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

COPY . .

CMD ["python", "app.py"]



  


  

  • Docker Compose :

  


  


  services:
  api:
    build: ./api
    ports:
      - "8000:8000"
    depends_on:
      - redis
      - mongo
    networks:
      - app_network
    volumes:
      - ./api:/app
    environment:
      - REDIS_HOST=redis
      - REDIS_PORT=6379
      - MONGO_URI=mongodb://mongo:27017/app_db

  frontend:
    build: ./frontend
    ports:
      - "3000:3000"
    depends_on:
      - api
    networks:
      - app_network
    volumes:
      - ./frontend:/app
      - /app/node_modules

redis:
    image: "redis:alpine"
    restart: always
    networks:
      - app_network
    volumes:
      - redis_data:/data

  mongo:
    image: "mongo:latest"
    restart: always
    networks:
      - app_network
    volumes:
      - mongo_data:/data/db

networks:
  app_network:
    driver: bridge

volumes:
  redis_data:
  mongo_data:



  


  Issue :

  


  When I try to access the /video endpoint, the following warnings appear :

  


  [ WARN:0@46.518] global cap_ffmpeg_impl.hpp:453 _opencv_ffmpeg_interrupt_callback Stream timeout triggered after 30037.268665 ms


  


  However, locally, the RTSP stream works fine using OpenCV with the same code.

  


  Additional Information :

  


  

  1. Network : The Docker container can successfully ping the camera IP (10.100.10.94).

  


  2. Local Video : I can read frames from a local video file without issues.

  


  3. RTSP Stream : I am able to access the RTSP stream directly using OpenCV locally, but not inside the Docker container.

  


  


  Code :

  


  Here's the relevant part of the code in my api/app.py :

  


  import cv2
from fastapi import FastAPI
from fastapi.responses import StreamingResponse

RTSP_URL = "rtsp://deneme:155115@10.100.10.94:554/axis-media/media.amp?adjustablelivestream=1&fps=10"

def generate_frames():
    cap = cv2.VideoCapture(RTSP_URL)
    if not cap.isOpened():
        print("Failed to connect to RTSP stream.")
        return

    while True:
        success, frame = cap.read()
        if not success:
            print("Failed to capture frame.")
            break

        _, buffer = cv2.imencode(".jpg", frame)
        frame_bytes = buffer.tobytes()

        yield (
            b"--frame\r\n" b"Content-Type: image/jpeg\r\n\r\n" + frame_bytes + b"\r\n"
        )

    cap.release()

@app.get("/video")
async def video_feed():
    """Return MJPEG stream to the browser."""
    return StreamingResponse(
        generate_frames(), media_type="multipart/x-mixed-replace; boundary=frame"
    )


  


  Has anyone faced similar issues or have suggestions on how to resolve this ?

  


  ---

  


• Gstreamer video increases latency with decreased FPS

  19 novembre 2024, par Ri Di

  I am using RPI 5 to stream the video :

  


  rpicam-vid -t 0 --camera 0 --nopreview --mode 2304:1296:10:P --codec yuv420 
           --width 640 --height 360 --framerate 10 --rotation 0 
           --autofocus-mode manual --inline --listen -o - | 
     ffmpeg -f rawvideo -pix_fmt yuv420p -s:v 640x360 -r 10 -i /dev/stdin 
            -c:v libx264 -preset ultrafast -tune zerolatency -maxrate 300k 
            -bufsize 50k -g 30000 -f mpegts tcp://192.168.0.147:1234


  


  View it with :

  


  gst-launch-1.0 -v tcpserversrc host=0.0.0.0 port=1234 ! queue ! 
    tsdemux ! h264parse ! avdec_h264 ! videorate ! video/x-raw,framerate=10/1 !  
    videoconvert ! autovideosink sync=false


  


  Problem is that with 10 FPS I get around 2s of latency ! While 56 or 120 FPS results in below 300ms latency.

  


  Is the problem in sender or reader side ? Or both ?

  


  I am not planning to use the 10 FPS, its only for demonstration of problem. But I would like to get lower latency at 56 FPS - just like at 120 FPS (around 80-100 ms difference) or maybe even better, as it seems to get lower with higher FPS.

  


  Maybe there is some kind of buffering parameter which holds frames ?

  


  (of course, when testing with higher FPS I change both numbers in sender and the one in reader command. The camera is v3 RPI official)

  


  Also I'd like to mention that same thing happens with ffplay :

  


  ffplay -i -probesize 3000 tcp://0.0.0.0:1234/?listen