extract frame from video in android device use JavaCV, but every time the model's inference results are not same

Hello, I am using the JavaCV to extract frame from video in android device and as input to my machine learning model, and I used the FFmpegFrameFilter to set the fps of video to get same number of frame every time. but every time I run the model, the model's inference results are not same, do you know the reason? these my code below :

       FFmpegLogCallback.set();
      grabber.start();
      FFmpegFrameFilter frameFilter = new FFmpegFrameFilter("fps=fps=5", "anull",
              grabber.getImageWidth(), grabber.getImageHeight(), grabber.getAudioChannels());
      frameFilter.setSampleFormat(grabber.getSampleFormat());
      frameFilter.setSampleRate(grabber.getSampleRate());
      frameFilter.setPixelFormat(grabber.getPixelFormat());
      frameFilter.setFrameRate(grabber.getFrameRate());
      frameFilter.start();
      Frame capturedFrame;
      Frame pullFrame;
      while (true) {
        try {
          capturedFrame = grabber.grab();
          if (capturedFrame == null) {
            Log.w("!!! end cvQueryFrame..video:{}", frontcamVideoSourcePath);
            break;
          }
          if (capturedFrame.image != null) {
            frameFilter.push(capturedFrame);
          }
          if ((pullFrame = frameFilter.pull()) != null) {
            if (pullFrame.image != null) {
              Bitmap bitmap = convertToBitmap.convert(pullFrame);
              setImage(bitmap, frontcamRunner.getSurface(), true);
              ResultProto objectResult = frontcamRunner.run();
              resultsObject.add(objectResult));

When I do the snapshot test, it shows every time the inference results are not same like:

Changed content at line 922:
expecting:
  ["scores: 0.9639015",
   "scores: 0.4762144",
   "scores: 0.29251003",
   "boxes: 359.04742",
   "boxes: 108.379776",
   "boxes: 366.3406",
   "boxes: 112.13722",
   "boxes: 359.66046",
   "boxes: 108.81177",
   "boxes: 366.08936",
   "boxes: 116.47825",
   "boxes: 358.7235",
   "boxes: 110.984955",
   "boxes: 366.55127",
   "boxes: 114.66323"]
but was:
  ["scores: 0.9536576",
   "scores: 0.47547817",
   "scores: 0.25218058",
   "boxes: 358.9765",
   "boxes: 108.3856",
   "boxes: 366.24115",
   "boxes: 112.15494",
   "boxes: 359.57233",
   "boxes: 108.810745",
   "boxes: 366.00732",
   "boxes: 116.48662",
   "boxes: 358.6304",
   "boxes: 110.98422",
   "boxes: 366.46204",
   "boxes: 114.66691"]

Changed content at line 941:
expecting:
  ["boxes: 358.75476",
   "boxes: 108.227554",
   "boxes: 365.989",
   "boxes: 111.96691"]
but was:
  ["boxes: 358.73907",
   "boxes: 108.22945",
   "boxes: 365.9668",
   "boxes: 111.97165"]

Changed content at line 949:
expecting:
  ["boxes: 358.5876",
   "boxes: 108.080986",
   "boxes: 365.78546",
   "boxes: 111.79181"]
but was:
  ["boxes: 358.58386",
   "boxes: 108.081505",
   "boxes: 365.78018",
   "boxes: 111.79303"]

Changed content at line 957:
expecting:
  ["boxes: 358.52716",
   "boxes: 107.90219",
   "boxes: 365.63968",
   "boxes: 111.619354"]
but was:
  ["boxes: 358.5263",
   "boxes: 107.902336",
   "boxes: 365.63843",
   "boxes: 111.619675"]

Changed content at line 965:
expecting:
  ["boxes: 358.5914",
   "boxes: 107.647606",
   "boxes: 365.6294",
   "boxes: 111.37422"]
but was:
  ["boxes: 358.5912",
   "boxes: 107.64763",
   "boxes: 365.62915",
   "boxes: 111.37428"]

Changed content at line 977:
expecting:
  ["boxes: 358.92163"]
but was:
  ["boxes: 358.92157"]

Changed content at line 979:
expecting:
  ["boxes: 365.94983",
   "boxes: 111.76167"]
but was:
  ["boxes: 365.94977",
   "boxes: 111.761696"]

Changed content at line 985:
expecting:
  ["boxes: 359.38873"]
but was:
  ["boxes: 359.3887"]

Changed content at line 987:
expecting:
  ["boxes: 365.98245",
   "boxes: 115.30535"]
but was:
  ["boxes: 365.98242",
   "boxes: 115.305374"]

Changed content at line 995:
expecting:
  ["boxes: 366.0174"]
but was:
  ["boxes: 366.01736"]