RealSense Linux (OpenCV?) error: VIDEOIO ERROR: V4L2: property frame_count is not supported

[breznak]

---

Required Info
Camera Model	{ D435 }
Firmware Version	05.12.01.00
Operating System & Version	Linux (Ubuntu 18.04)
Kernel Version (Linux Only)	5.3.0-24
Platform	PC/ntb
SDK Version	latest 2.31
Language	python/opencv
Segment

Issue Description

I'm trying to use the Intel RealSense2 camera (RGB) with YOLACT framework, please see dbolya/yolact#283

I'm getting the following crash:

$ python eval.py --trained_model=weights/yolact_base_54_800000.pth --score_threshold=0.15 --top_k=15 --video_multiframe=4 --video=0
Config not specified. Parsed yolact_base_config from the file name.

Loading model... Done.
VIDEOIO ERROR: V4L2: property frame_count is not supported
Initializing model... Done.
Traceback (most recent call last):
File "eval.py", line 1104, in <module>
evaluate(net, dataset)
File "eval.py", line 891, in evaluate
evalvideo(net, args.video)
File "eval.py", line 801, in evalvideo
active_frames = [{'value': extract_frame(first_batch, i), 'idx': 0} for i in range(len(first_batch[0]))]
File "eval.py", line 801, in <listcomp>
active_frames = [{'value': extract_frame(first_batch, i), 'idx': 0} for i in range(len(first_batch[0]))]
File "eval.py", line 790, in <lambda>
extract_frame = lambda x, i: (x[0][i] if x[1][i] is None else x[0][i].to(x[1][i]['detection']['box'].device), [x[1][i]])
TypeError: 'NoneType' object is not subscriptable

It seems to me

VIDEOIO ERROR: V4L2: property frame_count is not supported

would be a HW "error"/difference in the RealSense cameras? Is there a way to overcome this?
A related issue might be
#5628
, seems I would have to extract the frames in a loop from the HW manually, and feed it to the neural net. (?)

[ev-mp]

@breznak , the root cause has been identified and addressed by the framework's repo maintainers from the above link dbolya/yolact#283.
Closing as n/a.