2*RTX3090 , just 6FPS, why?

[zhaotun]

python demo.py --input-video test.mp4 --weights weights/jde_576x320_uncertainty.pt  --output-format video 

Namespace(cfg='cfg/yolov3_1088x608.cfg', conf_thres=0.5, input_video='test.mp4', iou_thres=0.5, min_box_area=200, nms_thres=0.4, output_format='video', output_root='results', track_buffer=30, weights='weights/jde_576x320_uncertainty.pt')

2021-10-19 17:11:49 [INFO]: Starting tracking...
Lenth of the video: 302 frames
2021-10-19 17:11:55 [INFO]: Processing frame 0 (100000.00 fps)
/home/kemove/disk_a/track/Towards-Realtime-MOT/Towards-Realtime-MOT-master/tracker/multitracker.py:142: NumbaWarning: 
Compilation is falling back to object mode WITH looplifting enabled because Function "tlbr_to_tlwh" failed type inference due to: non-precise type pyobject
During: typing of argument at /home/kemove/disk_a/track/Towards-Realtime-MOT/Towards-Realtime-MOT-master/tracker/multitracker.py (145)

File "tracker/multitracker.py", line 145:
    def tlbr_to_tlwh(tlbr):
        ret = np.asarray(tlbr).copy()
        ^

  @staticmethod
/home/kemove/anaconda3/lib/python3.8/site-packages/numba/core/object_mode_passes.py:177: NumbaWarning: Function "tlbr_to_tlwh" was compiled in object mode without forceobj=True.

File "tracker/multitracker.py", line 144:
    @jit
    def tlbr_to_tlwh(tlbr):
    ^

  warnings.warn(errors.NumbaWarning(warn_msg,
/home/kemove/anaconda3/lib/python3.8/site-packages/numba/core/object_mode_passes.py:187: NumbaDeprecationWarning: 
Fall-back from the nopython compilation path to the object mode compilation path has been detected, this is deprecated behaviour.

For more information visit https://numba.pydata.org/numba-doc/latest/reference/deprecation.html#deprecation-of-object-mode-fall-back-behaviour-when-using-jit

File "tracker/multitracker.py", line 144:
    @jit
    def tlbr_to_tlwh(tlbr):
    ^

  warnings.warn(errors.NumbaDeprecationWarning(msg,
/home/kemove/disk_a/track/Towards-Realtime-MOT/Towards-Realtime-MOT-master/tracker/multitracker.py:118: NumbaWarning: 
Compilation is falling back to object mode WITH looplifting enabled because Function "tlbr" failed type inference due to: non-precise type pyobject
During: typing of argument at /home/kemove/disk_a/track/Towards-Realtime-MOT/Towards-Realtime-MOT-master/tracker/multitracker.py (124)

File "tracker/multitracker.py", line 124:
    def tlbr(self):
        <source elided>
        """
        ret = self.tlwh.copy()
        ^

  @property
/home/kemove/anaconda3/lib/python3.8/site-packages/numba/core/object_mode_passes.py:177: NumbaWarning: Function "tlbr" was compiled in object mode without forceobj=True.

File "tracker/multitracker.py", line 120:
    @jit
    def tlbr(self):
    ^

  warnings.warn(errors.NumbaWarning(warn_msg,
/home/kemove/anaconda3/lib/python3.8/site-packages/numba/core/object_mode_passes.py:187: NumbaDeprecationWarning: 
Fall-back from the nopython compilation path to the object mode compilation path has been detected, this is deprecated behaviour.

For more information visit https://numba.pydata.org/numba-doc/latest/reference/deprecation.html#deprecation-of-object-mode-fall-back-behaviour-when-using-jit

File "tracker/multitracker.py", line 120:
    @jit
    def tlbr(self):
    ^

  warnings.warn(errors.NumbaDeprecationWarning(msg,
/home/kemove/disk_a/track/Towards-Realtime-MOT/Towards-Realtime-MOT-master/tracker/multitracker.py:105: NumbaWarning: 
Compilation is falling back to object mode WITH looplifting enabled because Function "tlwh" failed type inference due to: non-precise type pyobject
During: typing of argument at /home/kemove/disk_a/track/Towards-Realtime-MOT/Towards-Realtime-MOT-master/tracker/multitracker.py (111)

File "tracker/multitracker.py", line 111:
    def tlwh(self):
        <source elided>
        """
        if self.mean is None:
        ^

  @property
/home/kemove/anaconda3/lib/python3.8/site-packages/numba/core/object_mode_passes.py:177: NumbaWarning: Function "tlwh" was compiled in object mode without forceobj=True.

File "tracker/multitracker.py", line 107:
    @jit
    def tlwh(self):
    ^

  warnings.warn(errors.NumbaWarning(warn_msg,
/home/kemove/anaconda3/lib/python3.8/site-packages/numba/core/object_mode_passes.py:187: NumbaDeprecationWarning: 
Fall-back from the nopython compilation path to the object mode compilation path has been detected, this is deprecated behaviour.

For more information visit https://numba.pydata.org/numba-doc/latest/reference/deprecation.html#deprecation-of-object-mode-fall-back-behaviour-when-using-jit

File "tracker/multitracker.py", line 107:
    @jit
    def tlwh(self):
    ^

  warnings.warn(errors.NumbaDeprecationWarning(msg,
2021-10-19 17:12:00 [INFO]: Processing frame 20 (5.16 fps)
2021-10-19 17:12:02 [INFO]: Processing frame 40 (6.23 fps)
2021-10-19 17:12:05 [INFO]: Processing frame 60 (6.68 fps)
2021-10-19 17:12:08 [INFO]: Processing frame 80 (6.91 fps)
2021-10-19 17:12:11 [INFO]: Processing frame 100 (7.05 fps)
2021-10-19 17:12:14 [INFO]: Processing frame 120 (7.07 fps)
2021-10-19 17:12:17 [INFO]: Processing frame 140 (7.05 fps)
2021-10-19 17:12:21 [INFO]: Processing frame 160 (7.03 fps)
2021-10-19 17:12:24 [INFO]: Processing frame 180 (6.99 fps)
2021-10-19 17:12:27 [INFO]: Processing frame 200 (6.96 fps)
2021-10-19 17:12:30 [INFO]: Processing frame 220 (6.94 fps)
2021-10-19 17:12:34 [INFO]: Processing frame 240 (6.93 fps)
2021-10-19 17:12:37 [INFO]: Processing frame 260 (6.91 fps)
2021-10-19 17:12:40 [INFO]: Processing frame 280 (6.90 fps)
2021-10-19 17:12:44 [INFO]: Processing frame 300 (6.88 fps)
2021-10-19 17:12:44 [INFO]: save results to results/results.txt
sh: 1: ffmpeg: not found

[Qizec]

请问你解决这个问题了吗